special theme: smart things · 2019. 10. 21. · by gclaudi aeck er t, h d of h f raunhofe i ns t...

ERCIM NEWSNumber 119 October 2019ercim-news.ercim.eu

SmartThingsEverywhere

Special theme:

Joint ERCIM Actions

2

CONTENTS

SPECIAL ThEME

The special theme “Smart ThingsEverywhere” has been coordinated byMargarete Hälker-Küsters (FraunhoferAISEC) and Erwin Schoitsch (AIT)

6 Smart Things Everywhere -

Introduction to the Special Theme

by Margarete Hälker-Küsters (FraunhoferAISEC) and Erwin Schoitsch (AIT)

8 Blockchain and AI – Cyber-Physical

Production Systems, by PhilippSprenger and Dominik Sparer(Fraunhofer IML)

9 Product Quality Prediction in Batch

Processes with Small Sample Sizes

Using Siamese Networks, byChristian Kühnert, Johannes Sailer(Fraunhofer IOSB) and Patrick Weiß(Fraunhofer ICT)

10 Acoustic Quality Monitoring for

Smart Manufacturing, by SaraKepplinger (Fraunhofer IDMT)

12 Smart Pick-by-Light for Efficient

Storage and Production Processes,

by Hanna Herger and ThomasWindisch (Fraunhofer IIS)

13 Efficient Supply Chain Traceability

Using IoT Technologies, byAlexandros Fragkiadakis (FORTH),Theoharis Moysiadis and NikolaosZotos (Future Intelligence Ltd)

15 Components and Tools for Large

Scale, Complex Cyber-Physical

Systems Based on Industrial

Internet of Things Technologies, byApostolos P. Fournaris and ChristosKoulamas (ISI/RC ATHENA)

KEyNOTE

3 Cognitive is the New Smart

by Claudia Eckert, Head of theFraunhofer Institute for Applied andIntegrated Security AISEC

JOINT ERCIM ACTIONS

4 Simula Research Laboratory joins

ERCIM

5 2019 ERCIM Cor Baayen Young

Researcher Award for Ninon Burgos

and András Gilyén

16 Smart Municipality, by JenniferWolfgeher (FH Burgenland), MarioZsilak (Forschung Burgenland) andMarkus Tauber (FH Burgenland)

17 Teaching Sustainability and Energy

Efficiency with the GAIA Project

by Georgios Mylonas (Computer Tech-nology Institute & Press “Diophantus”)and Ioannis Chatzigiannakis (SapienzaUniversity of Rome)

19 Smart Intersections Improve Traffic

Flow and Road Safety, by MartinStriegel (Fraunhofer AISEC) andThomas Otto (Fraunhofer IVI)

20 Smart Solutions to Cope with Urban

Noise Pollution, by Jakob Abeßer andSara Kepplinger (Fraunhofer IDMT)

22 Transforming Everyday Life

through Ambient Intelligence byConstantine Stephanidis (FORTH-ICS)

23 Managing the Trade-off between

Security and Power Consumption

for Smart CPS-IoT Networks, byPatrizia Sailer (Forschung Burgenland),Christoph Schmittner (AIT) andMarkus Tauber (FH Burgenland)

25 ISaFe - Injecting Security Features

into Constrained Embedded Firmware

by Matthias Wenzl (Technikum Wien),Georg Merzdovnik and Edgar Weippl(SBA Research)

26 Enabling Smart Safe Behaviour

through Cooperative Risk Manage-

ment, by Rasmus Adler (FraunhoferIESE) and Patrik Feth (SICK AG)

28 Assessing the Quality of Smart

Objects, by Ivana Šenk (Inria andUniversity of Novi Sad) and JamesCrowley (Inria)

29 AutoHoney(I)IoT - Automated Device

Independent Honeypot Generation of

IoT and Industrial IoT Devices, byChristian Kudera, Georg Merzdovnikand Edgar Weippl (SBA Research)

30 Yogurt: A Programming Language

for the Internet of Things (IoT), byIvan H. Gorbanov, Jack Jansen andSteven Pemberton (CWI)

32 Smart End-to-end Massive IoT

Interoperability, Connectivity and

Security (SEMIoTICS), by NikolaosPetroulakis (FORTH), et al.

RESEARCh ANd INNOvATION

42 High Performance Software Defined

Storage for the Cloud, by ThomasLorünser (AIT), Stephan Krenn (AIT)and Roland Kammerer (Linbit HA-Solutions GmbH)

44 The Battle of the Video Codecs in

Healthcare, by Andreas S. Panayides(Sigint Solutions and Univ. of Cyprus),Marios S. Pattichis (Univ. of NewMexico) and Constantinos S. Pattichis(Univ. of Cyprus)

45 The VR4REHAB Interreg Project:

Five Hackathons for Virtual and

Augmented Rehabilitation, by DanieleSpoladore, Sonia Lorini and MarcoSacco (STIIMA-CNR, EUROVR)

ANNOuNCEMENTS

46 In Memory of Peter Inzelt

47 Dagstuhl Seminars and Perspectives

Workshops

47 HORIZON 2020 Project Management

47 Editorial Information

34 Towards Improved Mobility

Support in Wireless Sensor

Networks, by Amel Achour, LotfiGuedria and Christophe Ponsard(CETIC)

35 Designing IoT Architectures:

Learning from Massive Spacecraft

Telemetry Data Analytics, by OlivierParisot, Philippe Pinheiro and PatrikHitzelberger (Luxembourg Institute ofScience and Technology)

37 Autonomous Collaborative Wireless

Weather Stations: A Helping Hand

for Farmers, by Brandon Foubert andNathalie Mitton (Inria)

38 Automatic Detection of Allergic

Rhinitis, by Gregory Stainhaouer,Stelios Bakamidis and IoannisDologlou (RC ATHENA)

40 Supporting the Wellness at Work

and Productivity of Ageing

Employees in Industrial

Environments: The sustAGE

Project, by Maria Pateraki (FORTH-ICS) et al.

ERCIM NEWS 119 October 2019


imperative that solutions based on ultraprecise cognitivemodels can detect traffic situations in real time and trigger anappropriate response. This requires new solutions that pro-vide these cognitive abilities within cognitive sensors andcognitive edge components in vehicles and infrastructures.That way, these solutions will be able to respond veryquickly to the given situation, plan proactively and takeactions that are coordinated with the components in theimmediate surroundings.

Fraunhofer also develops speech-driven dialog systems witha special focus on domain-specific knowledge for applica-tion in various fields of business and industry. Using andcombining state-of-the-art components for speech recogni-tion, question/answering via knowledge graphs and speechsynthesis, our technologies in particular address the concretechallenges and needs of industries, enterprises and B2Bapplications. Moreover, these technologies “made inEurope” ensure technological sovereignty, data can be storedand processed within secure data spaces and the methods of“informed machine learning” make sure that the systems caneven be trained on small data sets.

Trustworthy electronics are also in our focus. They willbecome a key component especially for the implementationof products and solutions using artificial intelligence andmachine learning. Integrity can only be achieved if all partsof these systems - especially the electronic hardware (CPUs,SoCs, sensors, memory) as the basis of all software compo-nents based on them - are trustworthy. Backdoors andTrojans must be excluded. However, owing to the stronginternationalisation of various steps in electronics develop-ment, it is now difficult to guarantee this trustworthiness.Moreover, German and European companies today havehardly any alternatives to using electronic components fromuntrustworthy international sources if they want to defend orexpand their world market leadership with innovative, digi-tised, AI-based products. Within the cluster CCIT activitieshave been started to support a new Fraunhofer initiative thataims to build Trusted Electronics Platforms (TrEP).

The ability to share real-time information from varioussources in a controlled, i.e. data sovereign, and securemanner is a prerequisite for new business models and sus-tainable new value chains. Connected systems across com-pany boundaries with cognitive capabilities and highlysecure shared data spaces are key to maintain leading globalmarket positions and to benefit from digitisation and globalconnectivity. United research disciplines, as displayed in thecontributions of this journal, are correspondingly essentialfor the future.

[L1] https://www.cit.fraunhofer.de/

3

KEyNOTE

3

Cognitive is the New Smart

The age of smart devices and applications is not onlydawning, it is here. Every one of us interacts with connecteddevices multiple times a day, generating vast data to beanalysed across a range of industries, applications, and sce-narios. The number of objects equipped with a sensor andmeans for processing data is increasing exponentially.Correspondingly, the realisation of the “Internet of Things”is showing promising first results. For example, in telemedi-cine and patient monitoring, in the automotive industry, inbuilding automation and in smart home applications, inlogistics, to name only a few.

But we are not where we could be. Today’s internet-basedapplications are still focused on classical paradigms, like col-lecting and exchanging data between peers and processingvast amounts of data in centralised hubs by “hyper scaler”,which apply artificial intelligence (AI) to learn and to sup-port decision making. However, essential for the future is aninfrastructure that offers extended functions for knowledgegeneration. The internet of the future integrates technologieswhich emulate the cognitive abilities of humans – our per-ceptions as informed by all the senses; our awareness, imagi-nation, and memory; our ability to plan, to orient ourselves,and to learn. It forms a network of cognitive technologies,and thereby is becoming a cognitive internet. The cognitiveinternet provides intercompany platforms to merge data froma wide variety of sources for it to be accessed in a controlledway. What’s more, AI methods are integrated at the edge, e.g.right in the sensors. This has various benefits, such asallowing knowledge to be generated and used locally in realtime, and GDPR compliant processes to be supported byreducing the amount of data that must be stored andprocessed outside the controlled edge devices.

Researchers from various disciplines have to interact in orderto bring forward key technologies from sensors to intelligentlearning methods in data processing and the cloud, for inte-grating these really “cognitive” abilities into the future inter-net. In order to respond to various industries’ demands to com-bine applied specialist knowledge for comprehensive needs,the Fraunhofer-Gesellschaft set up a research cluster combin-ing experts from different domains, to address these demands.The Cluster of Excellence Cognitive Internet TechnologiesCCIT [L1] launched its mission in 2018 with 13 FraunhoferInstitutes pooling their expertise in order to face the chal-lenges of digitalisation and develop new solutions for a “Cog-nitive Internet”. Cyber security and data protection play animportant role to protect industrial knowhow and to supportdata sovereignty but also benefitting from the opportunities ofdigitalisation. Moreover, data processed by cognitive comput-ing technologies must be trustworthy, otherwise the AI algo-rithms will produce, for example, inaccurate forecasts andwrong decisions based on manipulated or faked data.

The application scenarios for cognitive technologies arequite diverse, ranging from logistics to agile and mobileindustrial manufacturing to autonomous driving. A self-organising production line, for example, has to be able toidentify and locate components, machine parts, to be able toadapt its processing automatically to improve the productionprocess. For applications such as autonomous driving, it is

Prof. Claudia Eckert,

Director of the Cognitive

Internet Technologies CCIT

cluster of excellence and

Head of the Fraunhofer

Institute for Applied and

Integrated Security AISEC.

Joint ERCIM Actions

Simula Research

Laboratory joins ERCIM

Simula is an ICT research laboratory that pursues

important and fundamental problems of science and

engineering in order to drive progress that is of genuine

value to society. Closely integrated into the research is

the education of future scientists and the development of

commercial ventures.

Since its establishment in 2001, Simula has grown and nowcomprises three research units in Norway. Despite the geo-graphic spread, the research foci remain concentrated on fiveareas: cybersecurity (in Bergen), communication systemsand machine learning (in Oslo), and scientific computing andsoftware engineering (in Fornebu).

Scientific computing underpins much of today’s technolog-ical and societal advances. Simula develops new mathemat-ical tools to enable multi-scale simulation models, from thecellular to the organ level, to interrogate fundamental bio-medical questions. The large amount of data generated in this

era offers enormous opportunities, Simula develops newalgorithms for data-driven simulations and the efficient useof modern large-scale computers.

Software systems are becoming increasingly intelligent,dynamic, and unpredictable. Such complex systems requireinnovative validation and verification paradigms, in order tohandle the unpredictability of future AI-enabled systems.

As modern society becomes increasingly digitized, it is ofutmost importance that our communication infrastructuresare extremely robust. Simula’s researchers identify and inves-tigate issues with society’s telecommunications infrastructureand aim to provide new technology and solutions to addressthese issues. For instance, our networks must be both robustand flexible in order to support a wide range of applications,from automation to self-driving cars to the tactile Internet.

There is also a critical need today for new cryptographicsolutions, not only to secure the processing of sensitive dataon cloud computing platforms but also to protect againstfuture attacks from quantum computers. Simula’s researchon cyber security has two primary goals: to acquire a deepunderstanding of how to protect sensitive informationthrough the use of mathematical methods, and to designsecure and reliable communications and storage solutions.

Machine learning and data science touch upon all ofSimula’s research fields, either by contributing to the devel-opment of these fields, or by actively using machine learningas part of the research. As with the other research foci,Simula’s machine learning researchers conduct basicresearch in order to strengthen the link between theory andalgorithms, but always with an eye towards linking algo-rithms to high-impact applications that are of particular rele-vance to society.

An integral part of conducting excellent research is edu-cating and training tomorrow’s scientists and technologyexperts. Simula provides educational opportunities fromMaster’s through the postdoctoral levels and these activitiesare closely integrated with research and innovation. Simulapartners with a large number of degree-awarding institutions,both within Norway and internationally, allowing Simula tooffer a range of possibilities for research stays abroad.

Innovation activities are an inherent part of technologyresearch, as the conceptual work is applied to real life.Simula provides both pre-seed support for technology entre-preneurs, with free office space and relevant support serv-ices, as well as real financial investments for promising startups. By realizing the commercialization of research results,both via industrial partnerships and the creation of spin-offcompanies, Simula keeps its focus on solving problems thatadd value to society.

By joining ERCIM, Simula hopes to build a strongerEuropean network within our scientific fields.

Please contact:

Kyrre Lekve, Deputy Managing Director Simula ResearchLaboratory, [email protected]


Simula pursues important and fundamental problems of science and

engineering in order to drive progress that is of genuine value to society.

Closely integrated into the research is the education of future scientists

and the development of commercial ventures. Photos: Simula.


2019 ERCIM Cor Baayen

young Researcher Award

for Ninon Burgos

and András Gilyén

The 2019 competition for the ERCIM Cor Baayen YoungResearcher Award was highly competitive with 16 final nom-inees. A selection committee composed of members fromERCIM and Informatics Europe eventually awarded twooutstanding young researchers in recognition of the out-standing scientific quality of their research and the impact onscience and society they have already achieved in their shortcareer: Ninon Burgos (CNRS) and András Gilyén (Caltech)with an honorary mention to Paris Carbone (RISE). Theaward will be presented at the European Computer ScienceSummit (ECSS) in Rome on Tuesday, 29 October 2019.

Ninon Burgos, nominated by Anne Canteaut (Inria) is aresearcher in the ARAMIS Lab at the Brain and SpineInstitute (ICM) of the French CNRS in Paris. She joined theARAMIS Lab in 2017 when she was awarded a PRESTIGEPostdoctoral Research Fellowship (a European “MarieSklodowska-Curie” fellowship programme). She completedher PhD in 2016 at University College London in the Centrefor Medical Image Computing under the supervision ofSébastien Ourselin. In 2012 she obtained a MSc in

Biomedical Engineering from Imperial College London andan Engineering degree from the French Graduate School inElectrical Engineering and Computer Science (ENSEA). Herresearch currently focuses on the development of computa-tional imaging tools to improve the understanding and diag-nosis of dementia. The main focus of her PhD was the devel-opment of image synthesis algorithms for MR-based attenu-ation correction in hybrid PET/MR scanners, and fordetecting pathological abnormalities in the reconstructedPET data. During her first postdoctoral position at UCL, shedecided to tackle the problem of automatic MR-based radio-therapy treatment planning in the pelvic region by devel-oping a new framework combining segmentation and imagesynthesis. Ninon has already received high recognition fromthe research community. She got more than ten invitations togive presentations at external meetings, received three travelawards to attend major international conferences, a prize forher oral presentation at the PSMR 2015 conference in Italy,and a prize for the best publication in the European Journalof Medical Physics “Physica Medica” in 2017.

András Gilyén, nominated by Ronald de Wolf (CWI andUniversity of Amsterdam), is a researcher at Caltech (previ-ously at CWI until June 2019). András graduated on May 29,2019, at the University of Amsterdam with the distinction“cum laude”. His thesis is about quantum algorithms, andmore generally about the theoretical computer science (TCS)aspects of quantum computing, an interdisciplinary fieldcombining physics, computer science, and mathematics. Histhesis is based on a number of publications in the very bestconferences: three papers in STOC and FOCS, which are thetwo most prestigious annual conferences in TCS, and are itsanalogues of Nature and Science; and one paper in SODA,which is the world’s premier algorithms conference. Thethesis develops a new framework for quantum algorithms,called “quantum singular value transformation”. The work ofAndrás has already been quite influential in the field ofquantum computing: other researchers have realizedAndrás’s new framework is extremely versatile and pow-erful, and are seeking out his help to apply it to their ownproblems. He recently became a postdoctoral fellow at theworld-famous California Institute of Technology (Caltech),working with Profs. John Preskill, Fernando Brandão, andThomas Vidick. He has also been accepted to participate inthe very selective program on quantum computing in Spring2020 at the Simons Institute in Berkeley.

5

2019 COR BAAyEN AWARd

Winners (ex aequo):• Ninon Burgos (CNRS), nominated by Inria• András Gilyén (Caltech), nominated by CWI

Honorary mention:Paris Carbone (RISE), nominated by RISE-SICS

Finalists:• Mário Antunes (IT Aveiro), nominated by INESC-ID• Constantinos Costa (Univ. of Pittsburgh), nominated by

the Univ. of Cyprus• Pavlos Fafalios (FORTH-ICS), nominated by FORTH-ICS• Riccardo Guidotti (ISTI-CNR), nominated by ISTI-CNR• Lucca Hirschi (Inria), nominated by Inria• Csaba Kerepesi (SZTAKI), nominated by SZTAKI• Sebastian Lapuschkin (Fraunhofer HHI), nominated by

Fraunhofer Gesellschaft• Elena Mocanu (Univ. of Twente), nominated by CWI• Panagiotis Papadopoulos (Brave Software Inc.),

nominated by FORTH-ICS• Theofanis Raptis (IIT-CNR), nominated by IIT-CNR• Christof Weiß (Univ. Erlangen-Nürnberg), nominated by

Fraunhofer Gesellschaft• Naomi Woods (Univ. Of Jyväskylä), nominated by VTT• Marcin Wrochna (Univ. of Oxford) nominated by the

University of Warsaw

https://www.ercim.eu/human-capital/cor-baayen-award

Ninon Burgos (left) and András Gilyén, the two winners of the 2019

Cor Baayen Young Researcher Award.


Special Theme: Smart Things Everywhere

thiness, Intellectual Property Rights andGDPR (General Data ProtectionRegulation of the EC) compliance. Thiswill allow to maintain Europe’s leadingposition in digital industry and business.CCIT, the Fraunhofer Cluster ofExcellence »Cognitive InternetTechnologies«, is an excellent examplehow to achieve these goals by joiningforces in research and innovation.

On European level, particularly DGCONNECT, DG for CommunicationsNetworks, Content and Technology,who manages the EC “Digital Agenda”,the HORIZON Programmes, amongthem ECSEL, a Joint Undertaking andPPP (Public Private Partnershipsbetween EC, national funding authori-ties and industrial resp. research part-ners) have taken up these challenges intheir Work Programs, with researchprojects on Smart Manufacturing,Automated Driving, Smart Farming,Multi-Concern Assurance (Safety,Security, Performance and otherDependability properties), Smart Citiesand Homes, and many traversal projectsacross domains and challenges. The pil-lars of these programs for“Digitalization of European Industry”,are IoT (physical meets digital), BigData (value from knowledge) and AIand Autonomous Systems.

In the booklet “My agenda for Europe”of Ursula von der Leyen, the newPresident of the European Commission,a chapter is dedicated to “A Europe fitfor the digital age”. It focuses on AI, IoT,5G, and ethical and human implicationsof these technologies, empoweringpeople through education and skills, andon protecting ourselves with respect tothe risks of these technologies topicsthat are highlighted in the keynote ofthis issue of ERCIM News.

Highlights provided by the articlesacross domains, applications and chal-lenges can be summarised as follows:

Industrial manufacturing processesplace high demands on quality, effi-

Introduction to the Special Theme

Smart Things Everywhere

by Margarete Hälker-Küsters (Fraunhofer AISEC) and Erwin Schoitsch (AIT)

Market intelligence firm IDC predictsthat by 2025 each of us will be inter-acting with a smart device several thou-sand times each day. This technologicalchange will have considerable impactson our lives and lifestyles, and will beaccompanied by new challenges, oppor-tunities and risks. In this edition wefocus on smart applications in severaldomains, tackle a few technologicalaspects and deal with security andquality issues.

The basis is laid by the Internet of Things- IoT (interactions and communicationbetween humans and smart devices) andthe Industrial Internet of Things - IIoT(in industrial context, machine-to-machine communication). However,“smart things everywhere” is not just IoTor IIoT, or mobile phones – it meansintelligence, cognitive systems and tech-nology, machine learning and artificialintelligence, security, big data and cloudconnectivity, involving many domains ofeveryday life and a main driver for thedigital transformation of our world.

International standardisation is alsoreacting to these evolving developmentsand challenges, covering technical aswell ethical and societal issues.Technical aspects are handled in theJoint Technical Committee of ISO/IECJTC1, Subcommittee SC41 (“Internet ofthings and related technologies”) andSC42 (“Artificial intelligence”). ForIoT, the main topics are: architecture;interoperability (a key issue that “mas-sively deployed systems work”); appli-cations and use cases; coordinationinvolving liaison or study groups on IoTtrustworthiness (a key issue for publicacceptance and liability); IIoT; societaland human factors; and blockchain secu-rity under IoT restricted resources. In thearea of AI, key activities focus on aframework for AI systems usingmachine learning and other foundationalstandards; big data; and particularly ontrustworthiness of AI. This covers manyimportant issues for the applicability ofAI components and systems in criticalapplications, like risk management,

interoperability, bias in AI systems andAI aided decision making, robustness ofneural networks or algorithms, but alsogovernance and ethical and societal con-cerns overviews.

The European Commission as well aslarge organisations, e.g. in context ofautomated driving regulations and rec-ommendations, have set up documentsand guidelines on ethical considerationsand trustworthiness of AI and comput-erised decision making. The mostimportant examples are the EC HighLevel Expert Group’s (HLEG) Report“Ethics Guidelines for Trustworthy AI”[L1], the report of Informatics Europeand ACM Europe “When ComputersDecide: European Recommendations onMachine-Learned Automated DecisionMaking”[L2] , and the IEEE initiativeon “Ethical aligned design” (even plan-ning certification activities) [L3].

These considerations on internationalactivities should complement the arti-cles reporting about applications in var-ious domains and on quality, safety,security and risk management in generalof such widely deployed systems of“smart things”. The articles are groupedinto five chapters:

• Smart Industrial Applications • Smart Cities, Buildings and Homes • Quality, Safety, Security and Risk

Management • Smart Things Networks and Plat-

forms • Other Applications of “Smart Things

Everywhere”.

The keynote “Cognitive is the newSmart” by Prof. Claudia Eckert, Directorof the Cognitive Internet TechnologiesCCIT cluster of excellence and Head ofthe Fraunhofer Institute for Applied andIntegrated Security AISEC, shows us theway beyond the current conventionaluse of Internet, smart devices and IoT,towards a “Cognitive Internet”. This is anetwork of cognitive technologies forknowledge generation and sustainablevalue chains, while preserving trustwor-

ERCIM NEWS 119 October 2019 7

ciency and productivity to stay competi-tive. In order to reach this goal, machinelearning (ML) and artificial intelligence(AI) methods are used for a wide rangeof data analysis, process control and pro-duction steering. ML usually needs ahuge amount of data. In the productionenvironment there are situations whereonly a small amount of high quality datais available, e.g. during the commis-sioning phase. In this case ML providesmethods known as one-shot or few–shotlearning. Data in industrial processes areprovided by a variety of sensors, oneexample is the application of acousticmonitoring for quality assurance. Thedata from acoustic sensors can e.g.define the right time for preventivemaintenance.

In today’s digital world, large amount ofdata is collected from different sources.These data have to be analysed and visu-alised, and anomalies have to bedetected within a short timeframe.Alarms and alerts have to be parsed inreal time. Often self–learning algorithmsare used to solve these kinds of prob-lems. These new technologies will affectmany areas, including manufacturing,financial services, healthcare, logistics,information security and also the opera-tion of satellite constellations as well asspacecrafts.

In logistics one example is the exploita-tion of IoT technologies in supply chaintraceability. From a wide range ofdevices, such as RFID, NFC, barcodes,Bluetooth, sensors, different types ofdata like velocity, temperature,humidity, location can be read. Thesedata combined with ML algorithms canimprove the whole supply chain for thebenefit of all stakeholders. An efficientstorage and pick up system based onwireless IoT technology for smart com-missioning is another application field inthe area of logistics.

Artificial intelligence technologies canalso be applied to improve the safety ofroad traffic, to visualise noise pollutionor to support farmers in their daily work.

Cameras monitor an intersection andsend real-time and high-quality videosto a server where objects (pedestrians,cars, etc.) are identified and critical situ-ations might be foreseen by machinelearning technologies. This information

will be sent to road users who can reactaccordingly. Based on a distributedacoustic monitoring system noisesources, noise levels and their contribu-tion to the noise pollution at a specificlocation can be detected on a detailedlevel. Data from different weather sta-tions are collected and analysed in asmart way to provide accurate weatherforecasts (e.g. risk of freezing).

Knowledge–based medical diagnosisand therapy systems have been aroundsince the 80s. Today, more advancedtechniques enable further applicationareas, for example, remote diagnosis ofallergic rhinitis is possible by analysinga short phrase uttered on a mobilephone. In an industrial working environ-ment, cameras and sensors can provideinformation in a non-invasive way onuser activities and states in order todetect health risks.

Last but not least are the smart applica-tions we might have in the home in thefuture. The article “TransformingEveryday Life through AmbientIntelligence” gives insights into ourfuture home life.

Many of the aforementioned applicationsare based on a wide range of deviceswhich are connected via a network andexchange information. To design, con-figure, manage and maintain these kindsof networks is challenging in terms ofinteroperability, scalability, energy con-sumption, real-time data and security. Afew articles are related to this topic.

With these new technologies and devel-opments, further challenges and risks forsafety, security and privacy (users) anddata arise.

A key topic is security, which has dif-ferent facets. First, security by design isdesirable. “Security by design” refers tothe systematic collection and assessmentof security goals, threats and countermea-sures. Based on the results, a securesystem is designed and implemented.

Many of the embedded systems do nothave the same level of security featuresthat is common in standard operationsystems. This is because devices are tai-lored to the specific application. A newapproach, based on binary rewriting toretrofit already existing IoT systems, will

be investigated in order to make theembedded systems more resilient againstunauthorised access attempts. It is vitalfor smart systems to be protected againsthacker attacks as effectively as possible.

The reliability of information in very dif-ferent situations has to be ensured. Thisbecomes crucial if, for example, personaldata, contractual data, financial dataand/or different stakeholders areinvolved. One answer is offered byblockchain technology together with theIoT and artificial intelligence. Another isthe Arrowhead framework, which pro-vides a chain of trust via mechanisms likecertificates and secure on boarding. TheInternational Data Space Association isconducting work in this area.

Systems used in real life (e.g. au-tonomous driving cars or medical diag-nosis) have to guarantee safety, whichmeans very high quality standards.

A characteristic of artificial and self-learning systems is that they may havean unsupervised unpredictable behav-iour. Based on the criticality of the appli-cation, different quality assurance andtest concepts must be developed in orderto guarantee that the systems are reli-able. Standards and certifications haveto be evolved.

One approach to this is “cooperative riskmanagement”, where all smart devicesshare their information and act together.Another approach is to develop a hierar-chical model for qualities for smartobjects. Based on existing qualitymodels, new dimensions are added,which take into account the capabilitiesthat arise with smart objects.

Links:

[L1] https://kwz.me/hEH[L2] https://kwz.me/hEE[L3] https://kwz.me/hEK

Please contact:

Margarete Hälker-KüstersFraunhofer-AISEC, Germany [email protected]

Erwin SchoitschAIT Austrian Institute of Technology,[email protected]

Today’s “cyber-physical production sys-tems (CPPS)” are largely closed IoTapplications with very limited contrac-tual capabilities and interoperability.Integrating AI and IoT into productionsystems enables improvement of effi-ciency when it comes to finding optimaljob routines in self-organised systems.The missing link here is the integrationof business logic and management rules,which support contractual, administra-tive and financial processes to transferthe concept of CPPS to industrial opera-tions. For this reason, researchers atFraunhofer Institute for Material Flowand Logistics IML and TU DortmundUniversity are developing blockchain-based CPPS in the Innovationlab inDortmund [L1].

The blockchain provides a distributed,tamper-proof dataspace for contractualas well as event-based data. AI-equippedsoftware agents, which are researchingand evaluating different organisationalrules for better job-handling, serve asdesignated light nodes. This infrastruc-tural element allows for distribution ofnew rules to all relevant shop-floor com-ponents via the blockchain. This featurealso ensures reliability and trust for the

generated transactions, which can serveas the basis for smart contract deploy-ment [1]. The blockchain also keepstrack of all agent-based decisions anddistributes them within the network, sothey can be adapted for different cir-cumstances. At the same time, crypto-tokens enable autonomous micro-trans-actions and support monetising actionson the shop floor at low transactioncosts. This way, a pay-per-use modelcan be introduced, even if the softwareand hardware equipment is provided bydifferent stakeholders.

Simultaneously, the blockchain sup-ports an identity management conceptfor CPPS without any additional inter-mediaries [2]. This means that the inte-gration of the CPPS into theblockchain’s P2P-network providesevery agent with a unique digital iden-tity. This way, every transaction caneasily be tracked to the device that pro-duced it. Furthermore, agents can beintroduced to or removed from the net-work without having to modify thesystem itself, greatly increasing agilityand adaptability. The network itself pro-vides the necessary legitimation for dig-ital identities. These redundant servers,

secure communication channels anddigital identities assigned to all agentsystems increase cyber-security andcyber-resilience [2].

In addition to this, the integration ofblockchain-technology with AI-equipped agents supports decision-making processes by providing one uni-form, transparent data pool and iden-tical criteria for decision-making, effec-tively removing communication efforts(see Figure 1) [1]. By using smart con-tracts as enablers for autonomousprocess flows, the agent systems withinthe CPPS project can thus negotiatemanagement actions and payments viamicro-transactions more easily, whilethe blockchain-network itself isexpandable more quickly and withreduced effort [1].

The project is a cooperation betweenFraunhofer IML and TU DortmundUniversity, with all experiments beingconducted in the Innovationlab inDortmund. The researchers have dif-ferent expertise in self-organising sys-tems and want to investigate how theintegration of administrative businessprocesses and applications can be exe-



Blockchain and AI –

Cyber-Physical Production Systems

by Philipp Sprenger and Dominik Sparer (Fraunhofer IML)

A new joint project between Fraunhofer IML and TU Dortmund University combines blockchain, IoT

and artificial intelligence (AI). The demonstrator for “cyber-physical production systems (CPPS)”

integrates smart contracts and blockchain technology into a multi-agent system at shop-floor level.

The system covers negotiations, financial transactions and agile self-organisation while employing

only limited hardware resources in a near real-time environment.

Figure 1: CPPS-Demonstrator.

cuted successfully. The cooperationstarted in June 2019 and a proof of con-cept demonstration is planned forOctober 2019.

The CPPS-project is part of FraunhoferIML’s “silicon economy” vision. Withinthis scenario, logistics, as a majordriving force of extensive digitisation,is situated at the centre of platformeconomies. Inspired by the B2C appli-cations of Silicon Valley, the siliconeconomy is the B2B equivalent, con-necting artificial intelligence, globalinterconnectedness, data-based busi-ness models, agile financial flows aswell as smart contracting and distrib-uted ledger technologies all in one dig-ital habitat. Originally regarded as amere service industry, the entire logis-tics discipline is expressible via algo-rithms and can thus serve as a powerfullaunch pad for distributed artificial

intelligence and industrial applicationsof biological systems, for example forswarm intelligence. Moreover, becauseof its interface role, logistics alsodemonstrate ideal conditions for wide-spread implementation of those tech-nologies across various industries and,because of its network structure, acrossthe globe. The CPPS project is set tocontribute to this vision of the siliconeconomy to enable fast and independentmeans of operation at shop floor level.Like its biological role model, self-organising swarms, the project aims todevelop and propel equally skilledcyber-physical production systems forindustrial applications and lay thegroundwork for a large-scale introduc-tion at a later stage.

Link:

[L1] https://www.innovationlab-logistics.com/

References:

[1] K. Salah, et al.: “Blockchain forAI: Review and Open ResearchChallenges”, in: IEEE Access, vol.7, pp. 10127-10149, 2019.https://doi.org/10.1109/ACCESS.2018.2890507

[2] D. W. Kravitz, J. Cooper:“Securing user identity andtransactions symbiotically: IoTmeets Blockchain”, in: GlobalInternet of Things Summit(GIoTS), Geneva, Switzerland:IEEE, pp. 1-6, 2017.https://doi.org/10.1109/GIOTS.2017.8016280

Please contact:

Philipp Sprenger, Dominik SparerFraunhofer Institute for Material Flowand Logistics, [email protected],[email protected]


Production processes, especially in man-ufacturing, must meet high standards forquality, efficiency and productivity inorder to remain competitive. Owing toincreasing process complexity and fre-quent plant conversions, however, pro-duction is often not performed effi-ciently. One way to increase plant effi-ciency is a systematic data evaluation,including data management and analysisof historical and streaming process data.For data analysis, machine learning(ML) and artificial intelligence (AI)have undergone rapid development andare currently at the centre of technicalinnovations. In applications with verylarge amounts of data available, e.g. inspeech recognition or image analysis,deep neural networks are currently thestate of the art.

In contrast to image analysis or speechrecognition, the use of machine learningmethods to analyse production data is alittle more complex because even exten-

sive production data often containscomparatively little information. Forexample, in a plant configured for serialproduction, process parameters are usu-ally kept constant, leading to the pointthat always the same data set is deliv-ered. On the other hand, data with ahigh information content but in smallquantities, is generated during commis-sioning or after major modifications ofthe plant. An example is the use of newraw material where parameters orprocess control strategies have to bechanged and adapted several times toachieve the desired results.

Hence, one current research topic in theFraunhofer Cluster of ExcellenceMachine Learning is to make MLmethods applicable for this type of data.ML methods, suitable for small datasets, are referred to as one-shot or few-shot learning and were first presented in2006 [1]. Recent examples where one-shot learning showed good results are in

the area of classifying images [2] anddetecting the road profile forautonomous cars [3].

In a current use case, a method calledSiamese Network [2] is used to predictthe product quality of insulation panelsproduced on a foam moulding machine.A Siamese network, sketched in figure1, is a neural network architecture inwhich two convolutional neural net-works with the same weights work inparallel. Each network takes a differentinput vector, in this case the processdata from two different productions,returning two comparable output vec-tors. From these two output vectors adistance measure is calculated, makingSiamese networks feasible for classifi-cation tasks if one of the two input vec-tors is labelled. To validate the perform-ance of Siamese networks on this typeof production data, different recipeswere carried out on the mouldingmachine and the quality of the resulting

Product Quality Prediction in Batch Processes

with Small Sample Sizes using Siamese Networks

by Christian Kühnert (Fraunhofer IOSB), Johannes Sailer (Fraunhofer IOSB) and Patrick Weiß (Fraunhofer ICT)

Deep Learning algorithms usually need a large amount of data. Still, when analysing measurements from

manufacturing processes, informative data in sufficient quantities is rather rare, making the task more

complex. Therefore, the development of so-called few shot learning algorithms, focusing especially on the

analysis of small data sets, is one of the current research topics of the Fraunhofer Cluster of Excellence.

plates was measured. The resultinginsulation plates were evaluated interms of their welding degree, compres-sive strength and bending strength andseparated into five classes: (1) Goodquality, (2) brittle, (3) burned, (4) bentand (5) low compression. In summary,130 experiments were conducted.

As a first step, the Siamese network wastested on the binary problem meaning toclassify if a plate is of good or badquality. For comparison and to validatethe performance, a K-NearestNeighbour algorithm (KNN) was takenas a baseline. Results showed that forthe one-shot case, meaning only onegood and one bad production was usedfor training, the Siamese network

achieved on the average an accuracy of83.2 % while the KNN ended up on80.1 %, meaning the Siamese networkoutperformed the KNN by about 3 %.When taking the complete data set witha separation of 70 % training and 30 %test data, on average the network endedup with a 97.6 % classification accu-racy, outperforming the KNN by about2%.

For the classification of five classes, thenetwork achieved a decent 52.3 % accu-racy for the one-shot problem (keep inmind that 20 % chance would be arandom guess) and finishing on anaccuracy of 88.7 % for the whole dataset. As for the binary case, KNN wasoutperformed by around 3 %.

In general, the results showed that evenfor small data sets ML methods areappropriate to analyse process data.Specifically, Siamese networks and theK-Nearest-Neighbour approach alreadyshow good results for one-shot learningin the classification of isolation platesfor quality control. As expected, in afew-shot scenario, with more data avail-able, the classification accuracy couldbe substantially increased. In all casesthe Siamese network led to a better per-formance compared to the KNN.

References:

[1] L. Fei-Fei, et.al.: “One-shot learn-ing of object categories”, IEEEtransactions on pattern analysis andmachine intelligence, 2006

[2] G. Koch, et.al.: “Siamese neuralnetworks for one-shot image recog-nition.” ICML Deep LearningWorkshop, 2015

[3] L. Huafeng et. al.: “DeepRepresentation Learning for RoadDetection through SiameseNetwork”, Computer Vision andPattern Recognition, 2019

Please contact:

Christian Kühnert Fraunhofer IOSB, [email protected]



Figure 1: rincipal design of a Siamese Network. Both convolutional neural networks share the

same weights. Process data from a product with unknown quality is fed into one network and

compared to a product with known quality. Depending on a defined distance measure (e.g. L2-

norm) the similarity of the two products is calculated and a prediction of the quality can be made.

Sensor technology already plays amajor role in the industrial environ-ment, and systems relying on machinelearning (ML), sensor combinations,and data from all levels to improve pro-duction processes, are constantlyevolving [L1]. However, the potential touse information gleaned from acousticemissions has not yet been investigatedin the context of industry. This is whereour in-house demonstrator, an airhockey table, comes into play, enablingus to show vividly how we can recog-nise different material properties (in thiscase, different kinds of pucks) by theiracoustic fingerprints. Players strike asmall plastic disc (the puck), which

moves on a small air cushion, from oneside of the table to the other. With thehelp of measuring microphones and asupervised learning approach, it is pos-sible to “listen” to which puck is beingplayed (Figure 1).

The main advantage of our approach isthe non-destructive quality control thatdoes not interfere with the (production)process or demolish the inspectedobject. It combines various measure-ment and analysis steps: precise soundrecording, pre-filtering of noise anduseful sound, as well as intelligentsignal analysis and evaluation usingML. Several steps are required prior to

recording the data, including: deter-mining how data will be acquired (i.e.,type of microphone, position of sen-sors, description of the context, defini-tion of recording length and method,sampling rate, resolution); context spe-cific synchronisation with additionaland other (sensor) data, and the defini-tion of an annotation policy.

Once these requirements are defined,the actual recording of data consid-ering different conditions for the laterclassification training takes place. Inour case, the conditions includeaudible differentiation of three pucksof different weights (18/19/23 grams)

Acoustic Quality Monitoring for Smart Manufacturing

by Sara Kepplinger (Fraunhofer IDMT)

“I describe this sound as ‘pling’ and the other as a ‘plong’…’”, the participant said, distinguishing between

hockey pucks made from different materials. Similarly, it is possible to “listen” to acoustic quality in the context

of industry. Based on a demonstrator using an air hockey table, we show an approach for acoustic quality

monitoring, applicable to smart manufacturing processes.


and materials (3D printed (material:PA2200)/3D printed with rubber ring(material: PA2200)/hard plastic orig-inal), as well as different kinds of (sim-ulated) background noises and playerbehaviour. We record continuouslyusing an USB audio interface (into amini PC in wave format using twodirectional microphones, one for eachside of the air hockey field), with 44.1kHz and 32 Bit. After the training thesystem based on various predefinedconditions, the ad-hoc analysis andinterpretation of data takes place. Theclassification starts during power-up ofthe air hockey table and is able todetect the different kinds of pucks onthe field.

In an industrial environment, the instal-lation of acoustic measurement tech-nology is similar: a measurementmicrophone picks up the characteristicsounds of a given process. Dependingon the application, we do the recordingwith several microphones or a micro-phone array. Algorithms of source sep-aration (as one possible solution amongothers) filter out ambient noises, so thatwe separate background noise fromuseful noise. For this purpose, we usemethods of source separation [1], origi-nally developed to separate individualinstruments from a music recording [2].

We use this approach, for example, inin-line quality assurance and predictivemaintenance scenarios for industrialuse cases. Here, the acoustic moni-toring system can provide additional,and even more precise, informationabout the quality and condition of prod-ucts or processes. This may be eitherintegrated into existing measurementsystems or used as a completely inde-pendent monitoring system.

The human auditory system is able todetect and distinguish individualsounds, even in noisy environments,and place them in specific contexts. Inindustrial manufacturing processes, itis often possible to make statementsabout the operating status of a compo-nent, engine or even an entire machineby listening attentively; experiencedmachine operators can hear problemsor errors. However, so far it is quite dif-ficult to detect these indications offaulty processes or products by anyother means [3]. Our basic premise isthat everything that is audible (andinterpretable, e.g. recognisable as a dif-

ference) is also measurable as an indi-cator for quality.

In order to reliably identify and auto-matically classify machine noises, theanalysis steps mentioned above areindispensable. Until now, extensivetraining data has been required to trainthe respective system reliably. This isone of the challenges when it comes tooptimising the recognition performancein practical use. One of the currentchallenges is the lack of availablemeasurement data, which is why, as afirst step, we have generated test databased on three application examples(electric engines, marble track, andbulk tubes) and made it available fortesting purposes [L2]. At this stage, wehave successfully tested our aforemen-tioned method in practice, togetherwith various industrial partners suc-cessfully, and reached the TechnologyReadiness Level (TRL) 6.

In the future, we would like to achievea high recognition rate with lesstraining data. Furthermore, we plan todevelop a self-learning system thatlearns from acoustic measurement datato assess the quality of products andprocesses. Additionally, we areworking to understand and interpretwhat acoustic information is heard bymachine operators and inspectors, andhow they perceive it. To this end, weare working on a parameterisation ofthe subjective factors.

Links:

[L1] J Walker, “Machine Learning inManufacturing – Present andFuture Use-Cases”https://emerj.com/ai-sector-overviews/machine-learning-in-manufacturing/.

[L2] Industrial Media ApplicationsDatasets 2019.https://www.idmt.fraunhofer.de/datasets.

References:

[1] E., Cano, et al.: “Exploring SoundSource Separation for AcousticCondition Monitoring in IndustrialScenarios”, EUSIPCO2017.

[2] J., Abeßer, et al.: “Acoustic SceneClassification by CombiningAutoencoder-Based DimensionalityReduction and ConvolutionalNeural Networks”, in: DCASE2017.

[3] S., Grollmisch, et al.: “SoundingIndustry: Challenges and Datasetsfor Industrial Sound Analysis(ISA)”, in Proc. of EUSIPCO2019.

Please contact:

Sara Kepplinger Fraunhofer Institute for Digital MediaTechnology IDMT, [email protected]

Figure 1: Air hockey table equipped with two measurement microphones recording the puck’s

sound (e.g., ‘pling’ and ‘plong’).

Wireless IoT technology for smartcommissioningThe Fraunhofer Institute for IntegratedCircuits (IIS) located in Nuremberg,concentrates on topics in the field ofpositioning and networking such as self-organising, wireless networking tech-nology, wireless data transmission andthe integration of sensor networks inproduction and logistics applications.

A technology developed by FraunhoferIIS for extremely energy-efficient radionetworking is s-net®. In recent years, ithas formed the basis for the interactionbetween information systems and thephysical world in increasing numbers ofprojects. In such cyber-physical systems(CPS), context-aware objects can recordand transmit their application and envi-ronmental situation and interact inde-pendently and self-organising with sev-eral users, via s-net®. The integration ofcontext-aware, networked objects is thekey for a continuous digitisation ofprocess-control and the Internet-of-Things (IoT).

Numerous IoT applications have alreadybeen implemented on the basis of s-net®networking technology. Especially inthe context of Industry 4.0 and SmartProduction, s-net opens up enormouspotential. The digitised production bene-fits from s-net®, inter alia, in the areasof supported assembly and intralogis-tics, e.g. smart picking systems.

Commissioning is often the central func-tion of warehouse logistics and has sig-nificant influence on other businessareas, such as production and assembly.Digitisation is also changing the require-ments for order picking: in addition toincreasing efficiency and reducing theerror rate, flexibility and mobility (forexample, when changing product ranges

or order fluctuations) are becomingincreasingly important criteria. Byusing wireless pick-by-light systems,these requirements can be met.

Wireless pick-by-light: flexibility formobile applicationsPick-by-light systems support the orderpicker with light signals for the intuitiveand quick location of a storage compart-ment. Specific information such as thewithdrawal quantity can be conve-niently and quickly indicated on the

shelf display - which results in a signifi-cantly higher picking performance. Atthe same time, a lower error rate isachieved and costly rework is reduced.However, common cable-connectedpick-by-light systems only partiallymeet the requirements of flexible andlow-effort picking. Fraunhofer IIS hastherefore developed a wireless pickingsystem based on the s-net networkingtechnology, which has the advantagesof previous pick-by-light systems andalso meets the requirements for flexi-bility and scalability.

With the aid of s-net technology, thepick-by-light system can be imple-mented wirelessly and mobile. Due tothe extremely high energy efficiency,the shelf displays can last up to severalyears battery-powered. The self-organ-ised communication between the indi-vidual compartment display nodes ofthe pick-by-light system makes it pos-sible to easily install the compartmentdisplay on the shelves, and facilitatestemporary storage structures and a rapidremodelling of removal compartments.

Optimisation of individualisedassembly processesA trend in the manufacturing industry,as in automobile production, is theincreasing individualisation of prod-ucts. Customer requirements need to berealised easily and cost-effectively up tobatch size 1. For manual assemblyprocesses, the complexity in terms ofquality and efficiency increases, sincedifferent components have to bearranged in different orders for eachproduct. This increases the demands onemployees in terms of ensuring qualityand efficiency of the assembly. Thewireless pick-by-light system allowsrequired objects, such as tools and cor-rect components, to be made available



Smart Pick-by-Light for Efficient Storage

and Production Processes

by Hanna Herger (Fraunhofer IIS) and Thomas Windisch (Fraunhofer IIS)

The requirements for efficient production and manufacturing processes are becoming increasingly complex in

global competition. Digitalisation should increase not only quality but also flexibility and the mobility of

processes. The solution for meeting these requirements can be “wireless IoT technologies”, which form a basis

for networking and interaction between machines (M2M) as well as between information systems and the

physical world. For IoT applications in the field of production and logistics, in particular, “wireless IoT

technologies” are increasingly used to digitise processes and make them more efficient. This article shows

how the field of commissioning is optimised by the wireless networking technology s-net®: Wireless pick-by-

light systems enable more efficient and flexible order-picking processes and support interactions with people.

Figure 1: Wireless order handling solutions

with the networking technology s-net® by

Fraunhofer IIS.

Figure 2: The self-organized radio communi-

cation between the individual compartment

displays of the pick-by-light system allows an

easy installation and is ideally suited for tem-

porary storage structures and rapid remodel-

ing of removal compartments and shelves.

in the right order directly at theassembly site.

TRILUM – Mobile Order HandlingSolutionThe wireless pick-by-light system [L3]based on s-net® technology is alreadyavailable as a commercial product“TRILUM” from the cooperationpartner AST-X GmbH. TRILUM guidesemployees to the correct storage posi-tion via LED light signal. Specificinformation such as the removal quan-tity can already be viewed on-site on ane-paper display in a convenient andtime-saving manner. With the aid of anacknowledgment button, the removal ofgoods can be confirmed quickly andeasily - the feedback to the warehousemanagement system is fully automaticand paperless.

TRILUM is mainly designed for use inlogistics and production and especiallysuitable for flexible, temporary andmobile storage structures. The system isparticularly easy to adapt, making itideal for environments where take-upcompartments and shelves are oftenredesigned or changed:• Order picking processes: Light-guid-

ed order picking optimises warehous-ing processes in logistics and intralo-gistics - cost-effectively and paper-less.

• Assembly operations: During assem-bly, required objects such as tools andcorrect components can be providedin the correct order.

• Distribution processes: Light-con-trolled distribution processes (put-to-light) ensure time-efficient, cost-effi-cient and error-free picking processes.

Links:

[L1] http://www.s-net-info.com/ [L2] https://trilum-solutions.com/ [L3] https://kwz.me/hEd

Reference:

[1] Hupp, Loidl & Windisch (2017),“Wireless IoT Für reinenBatteriebetrieb” in elektroniknet.de https://kwz.me/hEc

Please contact:

Hanna Herger, Thomas WindischFraunhofer Institute for IntegratedCircuits IIS, [email protected],[email protected]


The Internet of Things (IoT) has revolu-tionised the technological means thathave made feasible the interconnectionand interaction between the physical anddigital worlds. Billions of devices, world-wide, sense the physical world (ambienttemperature and light, humidity, weatherconditions, etc.), leading to the prolifera-tion of numerous applications and initia-tives including smart cities, e-health, pre-cision agriculture, etc. Another area thatIoT technologies can be successfullyexploited is supply chain (SC) trace-ability, where products are recorded asthey travel from the manufacturer to theconsumer. An IoT system can read datafrom a plethora of devices such as smarttags (RFIDs, NFC, Barcodes, BluetoothLow Energy), along with sensory datalike ambient temperature and humidity,vehicle speed, geolocation, and if intelli-gently combined together, can effectivelytrack the SC.

SC tracking offers numerous advantagesto all parties involved (producers,retailers, consumers): food safety

(quality deviation management, recallsin food crisis), perishable product pro-tection, origin verification and brandcertification, customer engagement andloyalty programs, monitoring, control,planning and optimisation of businessprocesses remotely [1]. Furthermore,the utilisation of IoT technologies canoffer low cost services, as the cost ofsensors and other related equipment(tags, etc.) significantly decreases overtime, owing mainly to the technologicaladvances of the hardware manufac-turing industry. At the same time, IoTtechnologies (communication proto-cols, interoperability standards,IoT/cloud architectures, security & pri-vacy algorithms) have substantiallymatured. For these reasons, IoT-basedSC tracking appears as an appealing andfeasible solution.

The aim of the Smart Product project[L1] is the design, implementation andevaluation of an IoT-based platform forSC tracking, product authenticity certi-fication and verification of origin, and

consumer engagement. The platformwill utilise IoT technologies at thedevice level (tags, sensors), as well assoftware and communication protocols,with emphasis on interoperability,energy-efficiency, security and privacy.This project aims to cover variousweaknesses of existing SC tracking sys-tems, such as: lack of monitoringbeyond retail, lack of transparency, lackof usability, flexibility and immediacy,fragmentation of technologies and lackof interoperability.

At operational level, the Smart Productproject will provide technological solu-tions by making the following feasible(Figure 1):• SC tracking by recording product-

related information such as storageconditions, geographical locations ofproduction, storage and sales. Thiswill make it possible for the produc-ers to monitor the products, partiallyaddressing the availability of prod-ucts in areas where for various rea-sons it is not allowed (e.g. exporting

Efficient Supply Chain Traceability

using IoT Technologies

by Alexandros Fragkiadakis (FORTH), Theoharis Moysiadis (Future Intelligence Ltd) and Nikolaos Zotos(Future Intelligence Ltd)

Traceability allows the identification and tracking of products as they travel through the supply

chain, from the manufacturer to the consumer. Smart Product [L1] is a research project aiming to

design and develop a secure IoT-based architecture for supply chain traceability, product origin

verification and authenticity certification.

smuggled products to other coun-tries), but will also enable consumersto be informed about the origin of theproducts and their storage conditionsduring transportation. Prerequisitesfor successful monitoring and record-ing are: the existence of a smart labelon the product or a package contain-ing a number of identical products,the existence of equipment for read-ing the smart labels, the existence ofequipment for measurement of stor-age conditions, such as ambient tem-perature and humidity sensors, and anappropriate database for data storing.

• Certification of products’ authenticityand origin, limiting the distribution ofcounterfeit products and therebyincreasing consumer confidence ingenuine products. The prerequisitesfor making this possible are: theintroduction of relevant productinformation into an appropriate data-base by certified users (producers,etc.), the existence of a smart label onthe product that indicates with appro-priate coding the identity of the prod-uct, and appropriate equipment forreading the smart label, such as asmart phone with the ability to readsuch labels.

• Stimulate consumer engagement withproducts through a smart phone appli-cation. Consumers, in addition to cer-tifying the authenticity of the prod-ucts, will be able to obtain further in-formation on the product in question,such as its ingredients, how to use it,etc. By creating appropriate user inter-faces on social networks, consumerswill be able to further engage withproducers/ marketers through appro-priate marketing methods.

The platform will be demonstrated andevaluated through a concrete use-case

(Figure 2) that employs producers/man-ufacturers and consumers, consisting ofthree steps: • Step#1: A producer/manufacturer

registers their product on the projectplatform, receiving a unique codethat identifies it (e.g. Electronic Prod-uct Code) that is further stored in theplatform’s database. It will also bepossible to state the monitoringparameters during product traceabili-ty (storage temperature, humidity,etc.). Product related informationsuch as optimum usage patterns,ingredients, geographical location ofproduction, expiry date and activa-tion of feedback to producers/manu-facturers will also be posted. In addi-tion, it will be determined which ofthe above information will also beavailable to the consumers.

• Step#2: The products are monitoredduring their shipment from the placeof manufacturing to that of disposal.Both in the storage and in the trans-port vehicles, there will be sensorsthat will record the monitoringparameters selected in Step#1. At the

same time, the unique code of thetransported products will be recordedand linked to the monitoring parame-ters while stored in the database.

• Step#3: The product is already on theshelf and steps # 1 and # 2 have beenfollowed. The product bears a smartlabel that is read by a smart phoneapplication that further communi-cates with the platform and collectsall available data about the productand imprints it on the mobile screenfor consumer information. Throughthe same application, consumers cansend feedback to the producer/manu-facturer, and can post on social net-works their experience with the spe-cific product.

The consortium consists of two part-ners, FORTH and the FutureIntelligence Ltd that receive fundingfrom the Operational Program of theRegion of Epirus, Greece.

Link:

[L1]: https://www.smartproduct.gr

Reference:

[1] C. Verdouw et al.: “Virtualizationof food supply chains with theinternet of things”, Journal of FoodEngineering, Elsevier, 2015.

Please contact:

Alexandros FragkiadakisICS-FORTH, Greece, [email protected]

Theoharis Moysiadis, FutureIntelligence Ltd, [email protected]

Nikolaos Zotos, Future IntelligenceLtd, Greece, [email protected]



Figure 1: Smart

Product IoT platform

conceptual view.

Figure 2:

Supply chain

tracking use-case.


In contrast to more restrained environ-ments, e.g. modern smart home automa-tion applications, in medium and large-scale installations, the costs of manualdevice configuration or substitution andreinstallation can easily reach unaccept-able figures. Examples of such a sce-nario can be, the typical deploymentprocedures to be followed in a smartenergy meter installation within abuilding where centralized IT infra-structure may not yet exist and forwhich different engineering teams maybe responsible for different phases ofthe installation.

In I3T, “Innovative Application ofIndustrial Internet of Things (IIoT) in

Smart Environments” project, severalobjectives and activities aim at the sim-plification of the aforementioned proce-dures by the design and implementationof the necessary components and toolsfor the deployment, configuration andmanagement of the industrial wirelessnetworked embedded systems existingat the base of the IIoT technologiesecosystem. We utilise the latest devel-opments in IETF standardisation activi-ties around the “thin-waist” of the IoT,being the IPv6-over-X/RPL/UDP/DTLS/CoAP related RFCs, and mainlythe results of the 6TiSCH and ACEWGs as well as the latest developmentin embedded system on chip devices(SoC) that consist of embedded proces-

sors and on chip FPGA fabric. Theactivities involve the study of thehighest achievable degree of the auto-matic initialisation and reconfigurationfor different application scenarios, byidentifying the different sets of parame-ters that need to be set off-line or on-line, and the limits between sets ofparameters initialised statically atdevice implementation, dynamicallyduring the deployment / commissioningtime, or dynamically during the normalsystem operation and the wholesystem’s lifetime. Reconfigurability isalso extended on the hardware level,where computationally demandingoperations are implemented as hard-ware components that are dynamically

Components and Tools for Large Scale,

Complex Cyber-Physical Systems Based

on Industrial Internet of Things Technologies

by Apostolos P. Fournaris and Christos Koulamas (ISI/Research Center ATHENA)

The application of Industrial Internet of Things (IIoT) technologies in large scale and complex cyber-

physical systems of systems (CPSoS), such as those on large, tertiary sector buildings, energy grid and

industrial production lines, still presents great challenges in the standardisation of the necessary

mechanisms, procedures, components and tools for the deployment, configuration, commissioning and

maintenance of their associated highly heterogeneous distributed subsystems. The problem is

magnified if one also considers the typical, resource constrained nature of most of the networked

embedded devices in a CPSoS, as well as the usually strict requirements in properties related to correct

and safe operations, such as real-time, reliable and secure processing and net working [2].

Figure 1: Heterogeneous IIoT technologies in reconfigurable Smart

Building and Smart Energy environments.

allocated on the FPGA fabric of moderncyber-physical system embeddedprocessors.

The I3T architecture is associated withthe network and end nodes of an IIoTinfrastructure. We adopt industrial net-works that use the TSCH operation ofIEEE 802.15.4e and structure on top ofit, a variation of the Zero-Touch SecureJoin protocol [1] that enables the securedeployment of a new node in an IIoTnetwork without user intervention. Wefurther enhance this mechanism so thatit can perform reconfiguration of theTSCH-6Top scheduling functions usingsecure CoAP messages. In parallel tothat, we introduce a hardware/softwareco-design mechanism inside each IIoTend node SoC so that each device cansupport dynamic reconfiguration of itshardware resources. More specifically,the designer has available a series ofhardware IP cores (with associated soft-ware drivers) that can be dynamicallydeployed after an application function-ality analysis on the IIoT device. The

analysis can highlight the computation-ally demanding operations that need tobe accelerated by hardware in order toretain the IIoT real-time responsivenessand safety requirements. Target exam-ples for hardware acceleration can besecurity/cryptographic primitive opera-tions, and machine learning processingon fast and large time series data fromthe monitoring of electrical grid criticalparameters or from machine and struc-tural elements vibration sensors [3].

As an outcome of the above activities,we created an integrated, medium scaledemonstrator for the smart energy andsmart building environment, whichincludes widely used, wired and wire-less, heterogeneous technologies of therelevant domains (e.g. BACNet, KNX,etc) which can also be attached to vir-tual environment simulators, in a hard-ware-in-the-loop fashion, capable ofdemonstrating the system operation at alarger scale. We have also designed anddeveloped an IIoT node prototype thatcan support the I3T dynamic hardware

accelerated reconfiguration and handlethe TSCH software network stack (withall the proposed dynamic, zero touchsecure I3T enhancements) [3].

Link:

[L1] https://i3t.isi.gr

References:

[1] M. Richardson: “6tisch Zero-TouchSecure Join protocol,” InternetEngineering Task Force Draft

[2] C. Koulamas et al.: “IoTcomponents for secure smartbuilding environments”, Springer,2017

[3] A. Fournaris et al.: “IntroducingHardware-Based Intelligence andReconfigurability on Industrial IoTEdge Nodes”, IEEE Design & Test,2019

Please contact:

Apostolos FournarisIndustrial Systems Institute / R.C.“Athena”, [email protected]



Increased digitisation and automatiza-tion require applications that make iteasy for citizens to report errors and beinformed about possible incidents intheir municipality. Such an incidentmanagement system can be citizen-based or automated via IoT (Internet ofThings). An interaction between a cit-izen and the local authority may be toreport an incident, e.g., an open man-hole. An automated (IoT supported) ver-sion of this interaction would be the col-lection of relevant data from sensors. Inany case, as actions are being triggered,both the citizen and the local authorityneeds information to be trustworthy.

Hence, on the municipality side, reliableand trustworthy data is essential, just asit is vital for citizens to know that theycan rely on information they receive -such as disaster warnings. Trust is the

most important requirement in emer-gency situations, in particular, but alsoin less extreme events like roadworks orconstruction areas. Existing smart cityapplications and citizen participationplatforms provide features relevant forsuch scenarios.

In the literature many platforms andapplications have been investigated[1],[2] on the basis of their features andpotential to enhance sustainability, butthe need for security and trustworthi-ness is rarely addressed. This isreflected in real-life citizen participa-tion and smart city platforms (e.g. plat-forms from Austria and Luxemburg,smart city projects from Stockholm andSingapore [L1-L4]). In these smartcities, the applications designed toincrease citizen participation and makeresidents lives easier, rarely guarantee

the reliability of exchanged informa-tion. Whilst including a number ofuseful features (like interactive maps,online participation in political discus-sions, waste service information andsmart parking), no existing system pro-vides incident management as an appli-cation in a trustworthy IoT environ-ment. To enable reliable incident reportsfrom citizens, information must becommunicated to the municipality in atrustworthy manner, allowing the localauthority to take action in dangerous sit-uations (e.g. open manholes). On thecitizen’s side, only approved informa-tion about actual conditions from themunicipality or sensors must bereceived, in order to ensure people’ssafety. Thus, a trustworthy frameworkto enable secure communication and tolink individual services from existingheterogeneous platforms is required.

Smart Municipality

by Jennifer Wolfgeher (FH Burgenland), Mario Zsilak (Forschung Burgenland) and Markus Tauber(FH Burgenland)

Digitalisation is already supporting individuals and smart cities in various ways. To increase

automatization and digitisation, decisions must be based on trustworthy information. We are

investigating the most common features of citizen participation and smart city platforms with

the aim of determining the trustworthiness of the digital environment in this context.


The Arrowhead framework [L5] is oneexample of a secure IoT framework thatcan integrate smartphones, sensors andexternal services in a smart city context.Sensors are already involved in theArrowhead framework in different con-texts, including industrial IoT (IIoT)and smart homes [L5]. This open-source framework provides many secu-rity functions by design, with the objec-tive to facilitate security, reliability,real-time communication and safety inlocal cloud automation. With the chainof trust principle, it enables the usage ofall services of the Arrowhead localcloud and the services of other cloudsthat are compliant with the Arrowheadframework, based on a secure on-boarding procedure.

In addition to the security advantages,Arrowhead supports a multi-cloud solu-tion that would make it possible to sticktogether the “patchwork” of applica-tions for smart city and citizen partici-pation projects, e.g. the many applica-tions of the smart nation of Singapore[L4] could be accessed via one service.The citizen could find the waste serviceas well as the parking service or taxservice in one place. The applicationscould be in one cloud or distributed inseparate, Arrowhead compliant, clouds,but the chain of trust would still beavailable from the citizen to eachservice.

The Arrowhead framework provides achain of trust via various mechanisms,

including certificates and secure on-boarding [3] to enable a trustworthyenvironment. Arrowhead is a fittingframework for the deployment of anincident report service in a trustworthyenvironment, capable of meeting theobjectives of security, reliability, real-time communication and safety.

Further research, in the EFRE project“civis 4.0 patria” (FE07) will includethe actual deployment of an incidentmanagement feature considering frame-works like Arrowhead to support thedevelopment of smart municipalities.

Links:

[L1] https://www.buergermeldungen.com[L2] https://kwz.me/hEN[L3] https://kwz.me/hEe[L4] https://www.smartnation.sg[L5] https://www.arrowhead.eu/

References:

[1] O. Gil, M. E. Cortes-Cediel, I.Cantador: “Citizen participationand the rise of digital mediaplatforms in smart governance andsmart cities”, Int. Journal of E-Planning Research (IJEPR), 8(1),19–34, 2019.

[2] A. M. Pozdniakova, et al.: “Smartsustainable cities: The concept andapproaches to measurement”, ActaInnovations, (22), 5–19, 2017.

[3] A. Bicaku, et al.: “Interacting withthe arrowhead local cloud: On-boarding procedure”, in 2018 IEEEIndustrial Cyber-Physical Systems(ICPS), pp. 743-748. IEEE, 2018.

Please contact:

Jennifer WolfgeherFH Burgenland, [email protected]

Teaching Sustainability and Energy Efficiency

with the GAIA Project

by Georgios Mylonas (Computer Technology Institute & Press “Diophantus”) and IoannisChatzigiannakis (Sapienza University of Rome)

Today’s students are the citizens of tomorrow, and they should have the skills and tools to

understand and respond to climate change. Green Awareness in Action (GAIA) has built an IoT

infrastructure within 25 schools in Europe, in order to enable lectures that target sustainability and

energy efficiency, based on data produced inside school buildings. The school community has

reacted very positively to this approach and has reduced energy consumption as a consequence.

GAIA [L1], a Horizon2020 EC-fundedproject, has developed a large-scale IoTinfrastructure in a number of schoolbuildings in Europe. Its primary aim isto raise awareness about energy con-sumption and sustainability, based onreal-world sensor data produced inside

the school buildings where studentsand teachers live and work.

Overall, 25 educational building sitesparticipated in GAIA, located inSweden, Italy and Greece. The IoTinfrastructure installed in these build-

ings monitors in real-time their powerconsumption, as well as several indoorand outdoor environmental parameters.However, this infrastructure would notbe particularly useful without a set oftools to allow access to the data pro-duced and provide functionality to sup-

Figure 1: Use case - incident report.

port educational activities. The GAIAChallenge [L2] is a playful interactiveplatform aimed at students, designed toserve as an introduction to power con-sumption and energy saving. In addi-tion, real-time data from sensors in thebuildings and participatory sensinghelp to visualise the real-life impact ofthe students’ behaviour and enablecompetitive gamification elementsamong different schools. The GAIABuilding Manager online applicationoffers visualisation of energy and envi-ronmental data.

The GAIA Challenge has been a suc-cess in GAIA’s software portfolio, withover 3,750 registered students andteachers using it as an introduction tosustainability and energy efficiencyconcepts. The end-user audience ofGAIA has largely comprised primaryand junior high school students, butsome high school and technical collegestudents as well. This variability in stu-dents’ age and origin has led to the useof a range of different approaches whenapplying GAIA’s tools within educa-tional activities.

The project completed its official activ-ities in May 2019. One of its biggestachievements was the construction ofan operational and reliable large-scaleIoT infrastructure within 25 schoolbuildings in Greece, Italy and Sweden.This infrastructure [2] includes over1,200 sensing endpoints, which havebeen based on open-source compo-nents. By combining the tools andproject methodology with data pro-

duced inside school buildings duringrelated activities in the schools, wehave energy savings of up to 15-20 %.

An important factor to encourageengagement is competition: studentswere intrigued by the prospect of com-peting with other schools and countries,and were further motivated to partici-pate in GAIA’s competitions for energysavings and related ideas. The projectalso held two official competitionsbetween schools during educationalyears 2017-18 and 2018-19, whichproved a valuable tool in motivatingand engaging the schools participatingin the project.

Following GAIA’s completion, the net-work of schools built during theproject, as well as the software toolsdeveloped, will continue to be active inthe following school year (2019-2020).The consortium members will collabo-rate with other research teams and con-tribute to the community in this fieldvia sharing datasets for experimentalevaluation. There are also methodolo-gies and educational activities availableon the project website [L1], offeringthe opportunity to replicate the resultsof the project to any school/teamwishing to do so.

This work has been supported by theEU research project “Green Awarenessin Action” (GAIA), funded by theEuropean Commission and theEASME under H2020 and contractnumber 696029.

Links:

[L1] http://gaia-project.eu/ [L2] https://gaia-challenge.com/

References:

[1] G. Mylonas, et al.: “EnablingEnergy Efficiency in Schools basedon IoT and Real-World Data”, inIEEE Pervasive Computing, Vol.17, Issue: 4, 2018,https://doi.org/10.1109/MPRV.2018.2873855

[2] D. Amaxilatis, et al.: “An IoT-based solution for monitoring afleet of educational buildingsfocusing on energy efficiency”, inMDPI Sensors, Special Issue inAdvances in Sensors forSustainable Smart Cities and SmartBuildings, 17(10): 2296,https://doi.org/10.3390/s17102296

[3] G. Mylonas, et al.: “An educationalIoT lab kit and tools for energyawareness in European schools”,Int. Journal of Child-ComputerInteraction, Vol. 20, 2019, pp. 43-53, ISSN 2212-8689,https://doi.org/10.1016/j.ijcci.2019.03.003.

Please contact:

Georgios MylonasComputer Technology Institute &Press “Diophantus”, [email protected]



Figure 1: Some of the basic elements in the GAIA approach (from left to right): an IoT node installed inside classroom measuring environmental

parameters, the world map of the GAIA Challenge on top of which students move between GAIA’s “missions”, and the ranking of different schools

and classes participating in the Challenge.


Increasing volumes of traffic are usingmunicipal road infrastructure, withsevere consequences for traffic efficiencyand the safety of road users. Vulnerableroads users (VRUs), such as pedestriansor cyclists, are involved in 46 % of lethalaccidents [1]. Exchanging informationbetween road users increases their per-ception and is thus a critical buildingblock to improve this situation.

The Smart Intersection, developed byFraunhofer Institutes IVI, AISEC, HHIand IIS in the IoT-COMMS project [L1]from 2018 to 2019, installs cameras attraffic junctions to monitor traffic.Those cameras send real-time and high-quality video to a road-side-unit (RSU),which detects and classifies objects suchas pedestrians or cars. Together withadditional information such as position,speed and direction of movement, theyare stored in a dynamic object map. Theobject map is then transferred from theRSU to cars via WLANp or LTE/5G-V2X.

This approach facilitates the cooperativeperception of surroundings, enablingroad users to utilise information fromother sensors. Thus, they can recogniseobstacles and other road users out ofplain sight, which prevents accidentsand allows more efficient traffic flows(Figure 1).

While modern cars can already utilisesensor-based object detection, para-metrisation and categorisation of objectsfrom within the moving car is chal-lenging. Shifting those tasks to roadinfrastructure, on the contrary, allowsreliable distinction between static anddynamic objects. Using spatially sta-tionary cameras permits “learning” thearea under surveillance, enabling earlydetection of critical situations and aquick reaction to them.

Previous approaches lack the bandwidthto transmit raw camera images between

infrastructure components and thustransfer only detected objects. Thislimits the effectiveness of the objectdetection and impedes the comparisonof detected objects, as each camera seesonly a particular part of the big picture.

DesignThe smart intersection, on the otherhand, utilises high-speed mmWavetransmission technology by FraunhoferHHI to transmit raw video frames frommultiple cameras to the RSU. There,object detection is performed using

background subtraction algorithmsdeveloped by Fraunhofer IVI.Subsequently, objects are classified, forexample “passenger car”, “cyclist” or“pedestrian” from multiview data. Todo so, Fraunhofer HHI employs convo-lutional neural networks. In addition,classification accuracy is improvedover time and trajectories can be calcu-lated, permitting tracing of temporarilyoccluded objects.

The key benefit of this centraliseddetection and classification approach isthat information from all cameras canbe used. Thus, the same object can beseen in multiple perspectives,improving the detection quality signifi-cantly and increasing the area that canbe perceived.

The object map is then written into astandardised collective perception mes-sage (CPM) and transferred from theRSU to cars via WLANp or LTE/5G-V2X. Thus, vehicles are enabled to

expand the range of their own sensorssignificantly and safety-critical situa-tions can be detected and controlled ear-lier or more safely at higher speeds.

Security and Privacy ConceptAnother key component of the SmartIntersection is the security concept forprotecting data against attackers. To pro-vide security by design, the modular riskassessment (MoRA) method of

Smart Intersections Improve Traffic Flow

and Road Safety

by Martin Striegel (Fraunhofer AISEC) and Thomas Otto (Fraunhofer IVI)

Smart intersections help to address increasing traffic density and improve road safety. By

leveraging data from infrastructure sensors, and combining and supplying those data to

road users, their perception can be improved. This aids in protecting vulnerable road users

(VRUs) and acts as a crucial building block for enabling automated and autonomous driving.

Figure 1: The smart intersection warns cars about the presence of pedestrians. This improves

the safety of pedestrians and other vulnerable road users.



Fraunhofer AISEC was applied [2].MoRA permits systematic collectionand assessment of security goals, threatsand countermeasures, resulting in aholistic and traceable security concept.

Human drivers and, in the near future,autonomous cars base decisions oninformation supplied by the smart inter-section, relying on the authenticity ofinformation. Thus, it is important toensure that the smart intersection cap-tures authentic information and that thedata has not been altered throughouttransmission and storage.

To ensure camera image authenticity,the RSU utilises a PRS Snapshot Sensorby Fraunhofer IIS, which capturestamper-proof Galileo PRS signals. Thisis used to apply unforgeable position-and time-stamps on the camera imagesand the generated object list. Public-keycryptography is used to sign andencrypt all messages during transmis-sion and storage. Tamper-proof hard-ware prevents physical attacks. Figure 2

shows the smart intersection and thesecurity concept.

To comply with data protection regula-tions, the centralised detection and clas-sification of objects permits the localisa-tion and anonymisation of numberplates and faces of pedestrians in cameraimages. However, authorities desire touse the raw, i.e. not anonymised, videodata captured by the smart intersectionto investigate accidents. Confidentialityand authenticity of those images isensured via cryptography as describedabove. Unlike standard camera surveil-lance systems, our solution stores allcamera data encrypted in a trust-storagehosted in a high-security facility. Thus,these data can only be accessed after ajudicial decision.

A proof-of-concept demonstration ofthe smart intersection has beendeployed at Fraunhofer IVI in Dresdenand will be used in further evaluation.We plan to enhance the intersectionwith further data-acquisition capabili-

ties, transforming it into a primarybuilding block of smart cities [L2].

Links:

[L1]: https://kwz.me/hEP[L2]: https://kwz.me/hE1

References:

[1]: European commission, „2017 roadsafety statistics: What is behind thefigures?”, 2018.[2]: Jörn Eichler and DanielAngermeier, “Modular risk assessmentfor the development of secureautomotive systems”, published at 31.VDI/VW - GemeinschaftstagungAutomotive Security, 2015

Please contact:

Martin StriegelFraunhofer AISEC, [email protected]

Thomas OttoFraunhofer IVI, [email protected]

Figure 2: Cameras

capture the smart

intersection and

provide data to

processing-units,

which send the data

over a secure link to

the road-side-unit

(RSU). The RSU

sends a map of

detected objects to

cars, extending their

perception range.

Smart Solutions to Cope with urban Noise Pollution

by Jakob Abeßer and Sara Kepplinger (Fraunhofer IDMT)

Noise pollution, especially in urban environments, can have negative health impacts. Smart city applications

for acoustic monitoring become essential to cope with the overall increasing noise pollution in residential

areas. Based on measurement data from a distributed acoustic sensor network, a web-based application

allowing for a real time visualisation of the citywide noise exposure was developed as part of the research

project “Stadtlärm”.

As described by Berg and Nathanson inthe encyclopaedia Britannica [L1]noise pollution is unwanted or exces-sive sound, which may derive fromindustrial facilities and other work-places, highway, railway, and airplane

traffic as well as from outdoor con-struction activities. The type of noisesource influences which noise protec-tion regulations are used. Therefore,one of our overall goals is to measurethe exposure of citizens to noise in dif-

ferent parts of the city based on theGerman technical guidelines for noisereduction (TA Lärm).

Within the research project“StadtLärm” (German for “city noise”)


[L2] a distributed acoustic monitoringsystem [1] was developed whose sensorunits are installed at the most relevantlocations of both sound emission andsound immission (see Figure 1). Figure1 gives an overview over the selectedtest site in the city of Jena, Germany.Acoustic sensors (orange squares) mon-itor the noise exposure at residentialareas (purple) and nearby sound emis-sion locations (red circles), such asopen-air venues, a soccer stadium, aswell as the most relevant transportroutes, like streets (green) and train plustram tracks (blue). Red arrows illustratethe most prominent sound propagationpaths towards nearby residential areas.In addition to localised sound levelmeasurements, the sensors automati-cally classify the most prominent soundevents. This makes it possible to detectwhich noise sources contribute most tothe noise pollution at a specific location.The system can provide valuable inputdata for systematic municipal planningin order to improve the residents’quality of life in the city.

The geographic location of the targetarea presents two main challenges.Owing to the valley-like elevation pro-file in Jena, most residential areas havehigher elevation than the monitoredsound emission locations. Furthermore,the prevailing westerly wind directionheavily influences the sound propaga-tion at the test site. The sensor unitscompute a set of standardised noiselevel parameters in near real-time with atemporal resolution of 125 ms. In addi-tion, a system for acoustic scene classi-fication based on a deep neural networkestimates the probability of nine dif-ferent acoustic scene classes once everysecond. The network uses a cascade ofconvolutional layers with intermediatepooling in order to recognise temporal-spectral patterns in short-term spectro-grams, which are characteristic of par-ticular sound sources [2].

Connected by a communication systembased on MQTT (Message QueueTelemetry Transport) [L3], these sen-sors communicate measurement data toa central server for data post-processingand storage. We implement the privacy-by-design approach by transmittingonly time-localised probabilities of dif-ferent acoustic scene classes.

The processing component (centralserver) publishes measurement results,

which include noise level measure-ments as well as acoustic scene classifi-cation results. Furthermore, it offersrequest/response interfaces forretrieving historical measurement data.This can be useful, for example, to cor-relate noise level measurements withpublic events that have taken place, inretrospect.

A web-based application receives andvisualises live measurement data at dif-ferent sensor locations on an interactivemap. The map also provides additionallayers of information about potentialnoise sources, such as road constructionsites, locations of bottle banks, andplaygrounds. It is possible to display thenoise level values measured at thesensor units in the map application inreal-time or aggregated over a config-urable period. For example, the user canget a condensed overview of the pastnoise situation by eliminating the inter-ferences caused by rush-hour traffic.

Finally, we will sketch several ideas forfuture improvements of the presented

sensor network. First, displaying andcommunicating basic measurementresults to the citizens via free andanonymous access to the applicationwill contribute to the public acceptanceof the system. In contrast, the citycouncil can receive an extended accountwith further internal administrativeinformation like conditions of theevents’ approvals and local residents’complaints. During the ongoing event,the regulatory agency of the city admin-istration can be informed automaticallyvia email if noise level limits areexceeded. Another approach to facili-tate the administrative decision proce-dure is to predict future noise situationsof planned events based on previouslymeasured data.

We are extending the proposed frame-work for further application scenarios inan urban environment. For instance, thesensor units along the main streetsallow traffic flow to be monitored andvehicle types to be identified.Furthermore, we will conduct long-termnoise level analyses to identify quietregions within the city. This informationcan be included in tourist recommenda-tions, e.g., recovery area suggestions, orused for solutions addressing wellbeingand health [3].

Links:

[L1]:https://www.britannica.com/science/noise-pollution[L2]: www.stadtlaerm.de[L3]: https://mosquitto.org/

References:

[1] J. Abeßer, et al.: “A DistributedSensor Network for MonitoringNoise Level and Noise Sources inUrban Environments,” in: Proc. ofFiCloud 2018.

[2] J. Abeßer, et al.: “Urban NoiseMonitoring in the Stadtlärm Project– A Field Report”, In: Proc. ofDCASE 2019.

[3] S. Kepplinger, et al.: “Perspectivesabout Personalization for mHealthSolutions against Noise Pollution”,in: Proc. of pHealth 2017.

Please contact:

Jakob Abeßer Fraunhofer Institute for Digital MediaTechnology IDMT, [email protected]

Figure 1: Targeted area in the city of Jena,

Germany. Orange squares represent acoustic

sensor units. Purple illustrates residential

areas. Red circles show example sound

emission locations. Streets are green and

train and tram tracks are blue. Red arrows

connect sound emission locations and nearby

residential areas.

ICS-FORTH has been investigating thepotential of AmI technologies indomestic life [L1]. Inside the“Intelligent Home” simulation spacelocated within its AmI Facility, everydayuser activities are enhanced with the useof innovative interaction techniques,artificial intelligence, smart objects,ambient applications, sophisticated mid-dleware, monitoring and decision-making mechanisms, and distributedmicro-services. This tightly-coupledconglomeration of software and hard-ware components seamlessly interop-erate so as to generate an ambient con-text-aware ecosystem that aims to: (i)improve the quality of life though appro-priate monitoring of health-related vari-ables (e.g. stress, sleep quality, nutri-tion); (ii) behave as an intelligent agentthat communicates with the users in anatural manner and assists them in theirdaily activities; (iii) transform the envi-ronment into an ambient notificationhub and personalised communicationcentre for the occupants (e.g. familymembers); (iv) implement a self-adap-tive, energy-efficient and eco-friendlyhome control middleware; and (v)enhance leisure activities by providingentertainment experiences.

In particular, the hardware facilities ofthe Intelligent Home consist of: (i) anextensive grid of sensors and actuatorsthat monitor and control various aspects(e.g. environmental conditions, energyand water consumption); (ii) smart com-mercial equipment (e.g. lights,speakers, locks, kitchen appliances, oildiffusers); (iii) smart devices and wear-ables (e.g. smart TVs, tablets, smart-phones, smart watches, smart wrist-bands of various sorts); and (iv) techno-logically augmented custom-made arte-facts. These components work in con-junction with the distributed computa-tional framework of the IntelligentHome, named AmIHomeOS, so as tofacilitate the realisation of comprehen-sive scenarios that accommodate var-ious use cases (e.g. a family of four, an

elderly couple, a disabled single adult)targeted to transform the domesticspace into an all-inclusive environmentthat assists end users in an intelligentand personalised manner.

The living room of the Intelligent Homehas been transformed into a smart spacethat exploits ambient technologies (i.e.

intelligent artefacts, technologically-augmented furniture) and services (e.g.video on demand, user profiling), asdepicted in Figure 1. In this context,AugmenTable (a smart coffee tableacting as a touch-enabled projectionarea, while being aware of the objectsplaced on it) and SmartSofa (a sofa

equipped with sensors that detect userpresence and posture and permit theexecution of actions via mid-air ges-tures) interoperate with AmITV (a hostof various interactive applications suchas movie player, news reader, musicplayer, etc.) and SurroundWall (aninteractive wall-based display) in orderto provide an enhanced viewing experi-

ence by transforming the overall spacearound the TV. Alongside, CaLmi aimsto reduce the inhabitants’ stress levels byreal-time monitoring of various psy-chophysiological (e.g. electrodermalactivity, heart rate) and contextual (e.g.workload, upcoming appointments,social life changes) variables and dis-



Transforming Everyday Life

through Ambient Intelligence

by Constantine Stephanidis (FORTH-ICS)

The ICS-FORTH Ambient Intelligence (AmI) Programme is a long-term horizontal interdisciplinary RTD

Programme aiming to develop pioneering human-centric intelligent technologies and environments which

seamlessly support everyday human activities and enhance well-being through human-technology symbiosis.

Figure 1: A photorealistic 3D rendering of the “Intelligent Livingroom”.

Figure 2: The CaLmi system in action.

tributing into the environment appro-priate relaxation programs that exploitthe existing ambient facilities (Figure 2).

The bedroom is generally considered asthe room of the house in which peopleretreat to unwind, relax, get ready andsleep. The intelligent bedroom supportsthese activities with the Hypnos frame-work, which aims to improve thequality of sleep by providing sleephygiene recommendations. This relieson various sensors integrated under thebed, and the inhabitants’ wearables tomonitor their physical activity (e.g.movement, time in bed), physiologicalsignals (e.g. respiration rate, heart rate,snoring) and various sleep-relatedparameters (e.g. time to fall asleep, timeasleep, sleep cycles) while resting.Moreover, Smart Wardrobe (a custom-made closet equipped with various sen-

sors and an embedded tablet) along withSmart Mirror (an interactive augmentedreality system) aim to provide outfitrecommendations based on contextualinformation (e.g. weather prediction,user’s daily schedule, user preferences,clothes availability), allow users to beimmersed in a “virtual mirror” wherethey can try them, and help them getdressed (i.e. helping them find whatthey are looking for).

Finally, the Intelligent Kitchen aims tointegrate ambient technologies to sup-port the inhabitants during the entirecooking process, ranging from the man-agement of the inventory and shoppinglist, meal selection and preparation, tocleaning-up. Currently, in the kitchen,smart appliances and custom artefacts(e.g. smart cupboards, smart food con-tainers) are accompanied by the

AmICounterTop and AmIBacksplashsystems, which constitute regular worksurfaces augmented with technology soas to be transformed into interactivedevices (e.g. primary or secondary dis-plays, kitchen scales), assist the abovetasks by communicating task-specificinformation to the user (e.g. highlightthe ingredients needed for the currentrecipe step, recommend recipes basedon goods about to expire, suggestbuying more milk when the last bottle isabout to be used).

Links:

[L1] http://ami.ics.forth.gr

Please contact:

Constantine StephanidisFORTH-ICS, Greece+30 2810 [email protected]


The fourth industrial revolution is basedon cyber-physical systems (CPS), wheremultiple components are interconnectedover the Internet of Things (IoT). Thesecomponents can communicate with eachother as well as measure and change theirphysical environment. Such applicationsoften require a high level of reliable andsecure data acquisition, but often havelimited energy resources. Thus, enablingpolicy-based adaptation to manage thetrade-off between e.g. security, reliabilityand resource usage will help developingsmart CPS. Supporting technologies mayinclude self-*, deep-learning, or neuralnetworks. In many applications it is cru-cial to maximise the lifetime of the com-ponents. Carrara et al. [1] have imple-mented an IoT-based management pro-gram to collect temperature andhumidity data. Tauber et al. [2] haveinvestigated energy efficiency and per-formance in a wireless-local area net-

work (WLAN) to identify upper andlower bounds of energy efficiency dueto different data flow characteristics.None of these have considered theimpact of different security settings onpower consumption, which is what wehave addressed in this paper. This willhelp to understand the magnitude ofpower saving due to different levels ofsecurity by e.g. smart-/self-adaptation ofthe application.

In order to investigate the electricalpower consumption under varyinglevels of security by different compo-nents of a CPS, we conducted measure-ments with a typical application, basedon service-oriented architecture (SOA).The measurement setup consists ofthree Raspberry Pi v3 with Rasbian[L1] as the operating system. The setupemulates a network in which data is col-lected with a sensor and is sent via

WLAN from client to server. The firstRaspberry is equipped with a GrovePi+[L2], which enables collecting data viaa “DHT22 temperature and humiditysensor” [L3]. These data are saved in alist and sent to an HTTP server via aWLAN access point (AP). The secondis the AP configured with HostAPD[L4], while the last one acts as theserver. Each Raspberry Pi has its ownVoltcraft Sem 6000 power plug (PP),which is responsible for measuring thepower consumption. An “expect script”[L5] was used to gather the data fromthese plugs. To send the data from theclient to the server, we implemented a“Spring Boot Rest Template” as aservice that sends the collected datausing a POST request. On the server wetracked the network communication viaWireshark and configured the followingencryption suites (ES) to indicate thedifferent power consumption:

Managing the Trade-off between Security and

Power Consumption for Smart CPS-IoT Networks

by Patrizia Sailer (Forschung Burgenland GmbH), Christoph Schmittner (AIT) and Markus Tauber(Fachhochschule Burgenland GmbH)

Making cyber-physical systems “smart” by managing the trade-off between security and resource usage

is of utmost importance for building sustainable industrial systems. For example, addressing cyber

security issues in such systems often require strong encryption. This may result in increased power

consumption on devices that often depend on limited energy supply. In this work, we present an initial

investigation into the usage of electrical power under different degrees of security in such situations to

understand and quantify the level of reduction of power usage due to varying degrees of security.

• ES1: TLS_ECDHE_RSA_WITH_AES_128

_CBC_SHA (Weak)


_CBC_SHA384 (Strong CBC)


_GCM_SHA384 (Strong GCM)

In our application we used TLS v1.2 asnetwork protocol and ECDHE_RSA askey exchange algorithm, which isdefined by a fixed ECDH keyexchange signed with an RSA certifi-cate [L6]. To illustrate the relationshipbetween power consumption andencryption suites (ES), we used dif-ferent key lengths for block encryptionand a message authentication algo-rithm. In the weaker cipher suite(ES1), we used the Block CipherAdvanced Encryption Standard (AES)with independent block and keylengths of 128 bits, as opposed to thestrong ones (ES2, ES3) with keylengths of 256 bits. Furthermore, theencryption options Cipher BlockChaining (CBC) and Galois/CounterMode (GCM) vary in performance dif-ferences due to algorithm and security[3]. As algorithm for message authenti-cation, we chose the Secure HashAlgorithm (SHA) with the differentkey lengths of 128 versus 384 bits.

With each encryption suite, we per-formed 21 test runs, each 25 minutesduration, with the collected data beingsent from the client to the server everythree seconds. To avoid latency or back-ground jobs from the compiler, we per-

formed a warm-up simulating a normaltest run.

As shown in Figure 2, the results of thetest runs show, that the stronger thecipher suite is, the more power is con-sumed by the client. The reason for thatis that the client is responsible for gener-ating the keys to communicate with theserver via the selected encryption suite.

This study, conducted as part of theEFRE project MIT 4.0 (FE02), showsthat the choice of data protection meas-ures via various cipher suites has a sig-nificant impact on power consumption:the stronger the level of security, thehigher the power consumption. Thus, it

supports our initial idea that a signifi-cant power usage reduction can beachieved by reducing the strength of thecypher suite – if the situation allows forit. In future work we plan to extendthose profiling activities as a basis forunderstanding the upper and lowerbounds of power usage for smart CPSwhich will be able to manage the trade-off between security, reliability andresource usage.

Links:

[L1] https://kwz.me/hEL[L2] https://kwz.me/hEl[L3] https://kwz.me/hEp[L4] https://kwz.me/hEg[L5] https://kwz.me/hEr[L6] https://kwz.me/hEY

References:

[1] M. Carrara, et al.: “An innovativesystem for vineyard management inSicily”, Journal of AgriculturalEngineering, 41(1), pp. 13-18, 2010.

[2] M. Tauber, S. N. Bhatti, Y. Yu:“Application Level Energy andPerformance Measurements in aWireless LAN”, IEEE/ACMGREENCOM 2011.

[3] Y. Hore, et al.: “BitstreamEncryption and Authenticationusing AES-GCM in DynamicallyReconfigurable Systems”,Advances in Information andComputer Security, pp 261-278,2008, ISBN: 978-3-540-89597-8.

Please contact:

Patrizia Sailer Forschung Burgenland GmbH, [email protected]



Figure 1: Architecture measurement setup. Three raspberries with power plugs (PP), which

send data collected with DHT22 sensor from client via WLAN connected through access point

to server. The power plugs measure the power consumption in milliwatts.

KEY LENGTH AES KEY LENGTH SHA ENCRYPTION OPTION

WEAK (ES1) 256 bits 384 bits CBC

STRONG CBC (ES2) 256 bits 384 bits CBC

STRONG GCM (ES3) 256 bits 384 bits GCM

Table 1: Overview differences used cipher suites.

Figure 2: Consumption of electrical power in milliwatts at the client. This study, conducted as

part of the EFRE project MiT4.0, shows that the choice of data protection measures via various

cipher suites has a significant impact on power consumption: the stronger the level of security,

the higher the power consumption. The issue of security needs to be further investigated and

discussed in the IoT area, taking electrical power savings into account. Understanding the

cooperation between these areas will make it possible to save more electrical power.


The Internet of Things (IoT) is formedby an ever-growing number of intercon-nected embedded systems. Anembedded system is a computing devicedesigned for a specific purpose thatoperates under certain constraints (e.g.,hardware and software tightly inter-twined to monitor traffic conditions on amotorway). Use-cases like smart cities,Industry 4.0, Car2X communication andambient assistive technologies promotenetworked embedded devices to becomeubiquitous companions in our dailylives. Unfortunately, many of the com-puting devices in the IoT lack even thesimplest form of security features,resulting in numerous successful attackson IoT devices in recent years. Securityfeatures, specifically exploit mitiga-tions, are procedures that reduce thelikelihood of an unpatched software vul-nerability being exploited by an adver-sary in such a way that it is possible tomislead the system to do unintendedoperations (e.g., address space layout

systems deployed to embedded com-puting platforms are configured toutilise security features. Additionally,many embedded operating systems lackthese kinds of features altogether. Thisis primarily caused by the fact thatmany embedded operating systems aimto support a wide variety of target hard-ware, including even the smallestdevices. For example, the widely usedFreeRTOS supports target platformsranging from Microchip’s SAMD20microcontroller with 16 kB non-volatileand 2 kB volatile memory to full fea-tured IA-32 based systems.Consequently, support for inherentsecurity features is omitted.Additionally, almost all security fea-tures introduced at an operating systemlevel protect either every component ofinterest (e.g., function entry point) ornone. This leads to an increase inresource consumption in terms ofmemory and runtime that might not bepracticable (e.g., the implementation of

ISaFe - Injecting Security Features

into Constrained Embedded Firmware

by Matthias Wenzl (Technikum Wien), Georg Merzdovnik (SBA Research) and Edgar Weippl (SBA Research)

The vast majority of the IoT is made up of computing devices that are highly specialised for their particular

purpose. Owing to their specialisation and the resulting constraints, such as energy consumption and the

deterministic fulfilment of deadlines (real-time requirements), these embedded systems tend not to share

many security features in common with standard operating systems. We aim to provide automated

approaches to implant security features into connected embedded systems to counter the lack of security

features in the backbone of the IoT and improve their resilience against unauthorised access attempts.

randomisation (ASLR) that is built intoevery Microsoft Windows operatingsystem derivate since Windows Vista(2007), as well as enabled by default inOpenBSD since 2003 and Linux since2005). The lack of security features inembedded systems is primarily basedon application specific constraints,hardware requirements (physical size,energy consumption, real-time require-ments [1]), software diversity (full fea-tures OS like Linux, *BSD, WindowsIoT Core, real-time OS like FreeRTOS,no OS at all) and security agnosticdesign decisions (e.g. unreflected usageof legacy code).

Well-known security features such asaddress space layout randomisation(ASLR), non-executable pages andstack layout transformation (SLX) arereadily available for general-purposeoperating systems. However, accordingto a study presented in 2017 [2], not allvariants of general-purpose operating

Figure 1:The core idea of project ISaFe read from left to right - Find locations within a firmware image that utilises data from external inputs in

certain ways. Then choose and implant selected security features to mitigate possible attacks on the detected locations in such a way that as much

as possible of the detected locations are protected and the firmware under investigations still adheres to its constraints (real-time, memory

consumption, etc).

a shadow-stack at an operating systemlevel doubles the amount of used stackmemory in terms of return addresses).Nevertheless, patching all availableembedded operating systems andalready circulating legacy systems atsource level is clearly illusive owing totheir large heterogeneity.

Therefore, the aim of the FFG-fundedproject ISaFe [L1] is to provide auto-mated approaches to implant securityfeatures into connected embedded sys-tems to counter the lack of security fea-tures in the backbone of the IoT startingwith September 2019. Together with theIoT startup Riddle & Code, whose aimis to provide an interface betweenembedded systems and blockchaintechnology, SBA-Research and the FHTechnikum Wien, all situated in Vienna,Austria, pursue a novel approach basedon binary rewriting to retrofit alreadyexisting IoT systems in order to makethem more resilient against unautho-rised access attempts.

Binary rewriting describes the alter-ation of a compiled and possibly(dynamically) linked program without

having the source code at hand in such away that the binary under investigationstays executable [3]. In the context ofhardening this means to implant secu-rity features into the binary under inves-tigation.

Unfortunately, constraints such as real-time requirements, or memory con-straints are very common in embeddedsystems software. Thus, a straightfor-ward approach like applying binaryrewriting for hardening purposes onevery point of interest (e.g., every func-tion call, every branch) in a binary withconstraints is not feasible due to theintroduction of possible constraint vio-lations. At the core, our approachrequires us to solve an optimisationproblem that maximises the number ofprotected vulnerable spots (locationswithin the firmware of interest thatmight be subject to attack, which weattempt to identify via taint tracking)with the most efficient security features,at the lowest possible costs - in terms ofmemory and run time overhead - whileobeying the system’s constraints. Theoverall idea if the presented approachcan be seen in Figure. 1.

Link:

[L1] https://kwz.me/hEB

References:

[1] H. Kopetz: “Real-Time Systems:Design Principles for DistributedEmbedded Applications”, secondedition, Springer, 2011.

[2] J. Wetzels: “Ghost in the Machine:Challenges in Embedded BinarySecurity”, Usenix Enigma, 2017.https://www.usenix.org/conference/enigma2017/conference-program/presentation/wetzels

[3] M. Wenzl, et al.: “From Hack toElaborate Technique - A Survey onBinary Rewriting”, 2019,https://doi.org/10.1145/3316415

Please contact:

Matthias WenzlTechnikum Wien, [email protected]

Georg Merzdovnik and Edgar Weippl SBA Research, [email protected],[email protected]



Smart machines need to behave safely ina smart way, predicting accidents andadapting their behaviour to avoid them,while efficiently fulfilling their originalmission. The prediction of possible acci-dents is a challenging task. It requiresperceiving the current situation andanticipating how it will evolve overtime. For safety reasons, any uncertaintyin this dynamic risk management mustbe compensated by a worst-caseassumption. This limits the performancesignificantly. For instance, if a smartmachine is not sure about the behaviourof another smart machine, it mustassume the most critical behaviour andbehave very cautiously even in situa-tions where it is not necessary.

A promising approach for maximisingperformance while ensuring safety iscooperative risk management. If allsmart machines, regardless of the man-ufacturer of the machine, shared theirinformation about the current situationand its evolution, we could minimisethe number of worst-case assumptions.However, the claims that each indi-vidual machine makes about the situa-tion and its evolution could be wrong.Thus, Fraunhofer IESE proposesextending these claims to a machine-readable assurance case capturing theunderlying reasoning and assumptions.

As shown in the left part of Figure 1, anassurance case is a reasoned and com-

pelling argument, supported by evi-dence such as test results, that an entityachieves its objectives and constraints.

As shown in the right part of Figure 1,the argument generally consists of sev-eral reasoning steps. Each step statesthat a higher-level claim is true if somesub-claims are true and some assump-tions hold.

The idea behind machine-readableassurance cases is to enable runtimeevaluations of the assumptions used inthe argumentation. A smart machine canevaluate if the assumptions used in theassurance case received from anothersmart machine hold in the current usage

Enabling Smart Safe Behaviour

through Cooperative Risk Management

by Rasmus Adler (Fraunhofer IESE) and Patrik Feth (SICK AG)

Machines in an Industry 4.0 context need to behave safely but smartly. This means that these smart

machines need to continue to precisely estimate the current risk and not shut down or degrade

unnecessarily. To enable smart safe behaviour, Fraunhofer IESE is developing new safety assurance

concepts. SICK supports related safety standards to implement these concepts in an industrial setting.


scenario. For instance, if the providedinformation is only correct if a tempera-ture is within a certain range, then asmart machine receiving this informa-tion can check if the temperature is cur-rently within this range. Furthermore, asmart machine may evaluate theassumptions with respect to the requiredintegrity. If there are too many arguableassumptions, then the smart machinemight decide that the integrity is insuffi-cient.

Fraunhofer IESE is developing solu-tions for implementing this idea in theDEIS project [L1]. In this project, weuse the term Digital DependabilityIdentity (DDI) for all the informationthat uniquely describes the depend-ability characteristics of a system orcomponent [1]. We chose SACM [L2]as the basis for describing DDIs,because SACM is the meta-model ofassurance case languages.

However, the concept of ensuring safetythrough machine-readable assurancecases is in conflict with traditional stan-dards in the domain of industrialautomation. In this domain, it iscommon to safeguard a critical applica-tion by using dedicated protectivedevices, mainly sensor devices with afixed data evaluation function, that pro-vide a certain safety function with aguaranteed level of integrity; e.g., asafety laser scanner detecting an intru-sion into a preconfigured safety field.These devices and their provided func-tions are conformant to harmonisedstandards such as IEC 61496-3. Suchstandards describe for a specific tech-

nology – for example in the case of IEC61496-3, active opto-electronic protec-tive devices responsive to diffusereflection – very precise requirementsfor achieving a particular integritylevel.

In European countries, the use of suchdevices is heavily stimulated as theMachine Directive explicitly coversprotective devices designed to detectthe presence of persons. Along with theexistence of specific harmonised stan-dards, this has fostered the belief in thedomain of industrial automation thatsafety equals conformance with suchstandards. This limits the possibilitiesfor building safe systems, as it con-strains the options for reducing the riskof a critical application to existing stan-dardized protective devices. In partic-ular, it hinders the implementation ofthe above-mentioned cooperative riskmanagement.

To overcome this limitation, SICK sup-ports the recently published technicalspecification IEC TS 62998-1. Thistechnical specification gives guidancefor the development of safety functionsbeyond existing sensor-specific stan-dards such as IEC 61496-3. It facilitatesthe use of new sensor technologies (e.g.,radar, ultrasonic sensors), new kinds ofsensor functions (e.g., classification ofobjects, position of an object), combina-tions of different sensor technologies ina sensor system, and usage under newconditions (e.g., outdoor applications).To this end, IEC TS 62998-1 requiresanalysing the capabilities of a sensorused in the intended application. A

system realising cooperative risk man-agement can be considered as a safety-related sensor system with a respectivesafety function in this new technicalspecification.

To support future Industry 4.0 applica-tions, sensor models could incorporatethe information required by IEC TS62998-1 and could become part of amachine-readable assurance case of thedevice. With this information stored inthe assurance case, it would becomepossible to use the device with its pro-vided sensor functions in critical appli-cations unforeseen during the initialdevelopment of the device.

Links:

[L1] http://www.deis-project.eu/home/ [L2] https://kwz.me/hEb

Reference:

[1] E. Armengaud et al.: “DEIS:Dependability EngineeringInnovation for Industrial CPS”, in:C. Zachäus, B. Müller, G. Meyer(eds): ‘Advanced Microsystems forAutomotive Applications 2017’,Lecture Notes in Mobility.Springer, 2018.

Please contact:

Rasmus AdlerFraunhofer Institute for ExperimentalSoftware Engineering IESE, Germany [email protected]

Patrik FethSICK AG, [email protected]

Figure 1: Illustration of an

assurance case.

What is a quality? Ideally, qualities aredistinctive attributes that can be used tocharacterise things or to provide a stan-dard for their comparison. The effectiveuse of a quality requires an establishedway to assess it through an objectivetest. Such tests may be quantitative, pro-viding a numerical value or qualitative,providing a symbolic label. In eithercase, the test must provide an objectivereference for comparison.

The use and measurement of qualitieshas been studied in a number of relateddomains, including manufacturing, soft-ware engineering, communications,human-computer interaction, multi-media systems and Internet of Things. Ingeneral, the study of qualities in each ofthese domains has given rise to hierar-chical conceptual frameworks, with theconcept of “quality” being broken downinto a variety of more detailed character-istics that can be evaluated with one ormore tests.

In many domains, quality dimensionsfollow similar principles. For example,in manufacturing Garvin [1] proposedeight critical dimensions of quality thatserve as a framework for analysis andproduct evaluation: performance, fea-tures, reliability, conformance, dura-bility, serviceability, aesthetics, and per-ceived quality. Similarly, in softwareengineering the scientific and industrialcommunities have developed a varietyof quality models, leading to the adop-tion of ISO/IEC standard 25010 [2],where quality is represented by twomajor aspects: product quality andquality in use. Product quality includesfunctional suitability, performance effi-ciency, compatibility, usability, relia-bility, security, maintainability, andportability. Quality in use includes effec-tiveness, efficiency, satisfaction,freedom from risk, and context cov-erage.

Similar patterns may be found in otherdomains, where we can find intrinsic

qualities such as performance, function-ality, conformance, correctness, relia-bility, efficiency, durability, reusability,maintainability, portability, compati-bility, security and privacy; as well asuser-related qualities such as usability,learnability, ease of use, naturalness ofinteraction, understandability, per-ceived quality, aesthetics, satisfaction,appropriateness, joy of use, usefulness,or trust.

Smart objects have emerged, as ordi-nary objects are increasingly aug-mented with digital technologies to pro-vide embedded functions for percep-tion, action, interaction and intelli-

gence. To build a comprehensivequality model for smart objects, we pro-pose to draw on the quality dimensionsidentified in related domains, and toidentify new quality dimensions thattake into account the capabilities thatarise with smart technologies. Such newcapabilities cover the features that makethe objects smart, such as connectivityand cognitive abilities. These includeabilities for sensing, actuating, interpre-

tation, communication and networking,as well as abilities for easy, natural, andconvenient interaction with people. Avariety of qualities are possible,although the relative importance mayvary with the type of smart objects.

We base our approach on different inter-action dimensions of smart objects(Figure 1). We start with the intrinsicdimension that addresses the self-ori-ented qualities of the smart object, andadd three extrinsic dimensions of inter-action that the smart objects have withthe environment, with other objects, andwith people. Accordingly, we propose ahierarchy of detailed quality character-

istics along the following main dimen-sions:

1. Innate qualities, related to the corefunction of the smart object. Theseinclude functionality and confor-mance aspects, as well as qualitiesthat relate to efficiency of the smartobject, such as energy management,reliability, and availability, and quali-ties that take into consideration the



Assessing the Quality of Smart Objects

by Ivana Šenk (Inria and University of Novi Sad) and James Crowley (Inria)

With the encroachment of smart objects into our lives, it is increasingly important to define a

principled approach to estimate the value of alternative technologies at design time, to define

product specifications and to compare similar products. In mature domains, such approaches are

based on properties that are referred to as “qualities”. We are working to develop a hierarchical

model for qualities for smart objects based on different modes of interactions.

Figure 1: Interaction

dimensions of smart

objects.

product lifecycle, such as durability,maintainability, portability, upgrade-ability, or recyclability.

2. Cognitive qualities, based on abilitiesto acquire, interpret and apply knowl-edge and skills. These include con-ceptual qualities that allow the smartobject to behave in a appropriatemanner, by perceiving, learning,understanding and reasoning aboutthe environment, and taking relevantactions.

3. Social qualities, based on abilities forcommunication, networking, andinter-object relationships. Theseinclude aspects of social interactionbetween smart objects, from the typesof connections that enable interactionto the possibilities that these connec-tions provide. They cover compos-ability through the diversity, capacity,compatibility and extendability ofconnections, availability of connec-tions through discoverability, accessi-bility and transitivity, and sociabilitythrough the ways of forming rela-tions, inter-object controllability andcollaborative ability.

4. Suitability qualities, which includethe capacity to act and interact withthe users in a manner that is appropri-ate for the task and context. To meas-ure suitability, we reflect upon theusability aspects such as learnability,ease of use, user control, understand-ability, or predictability, and accept-ability aspects, such as privacy, secu-rity, trust, aesthetics, engagement,satisfaction, or usefulness.

We are currently working to establishadequate objective ways to test andmeasure these proposed quality aspectsfor smart objects. Some of these aspectscan be tested under controlled labora-tory conditions, while the others requireevaluation under real world conditions.In order to assess our quality model weare applying the proposed tests to dif-ferent types of existing smart objects.Our objective is to promote these ideasas an approach for defining, describing,measuring and comparing smartobjects, with details to be refined anddeveloped by the larger community. Inaddition, as the boundary between

smart objects and smart services isincreasingly fluid, we believe that anynormative reference established forsmart objects should also be useful forsmart services including those providedby assemblies of smart objects.

References:

[1] D. Garvin: “Competing on theeight dimensions of quality,” Harv.Bus. Rev., 1987.

[2] “Systems and SoftwareEngineering - Systems andSoftware Quality Requirementsand Evaluation (SQuaRE) - Systemand Software Quality Models,”ISO/IEC 25010, 2011.

Please contact:

Ivana Šenk, Inria, France andUniversity of Novi Sad, [email protected]

James Crowley Inria, [email protected]


The interconnection of physical devices,vehicles, household appliances andother objects with electronics, software,sensors and actuators has become anintegral part of our modern lives. Theindustrial sector is also undergoing achange in device communication.Traditionally, automated factories andcritical infrastructure were strictly sepa-rated from the Internet. However, sincethe advent of Industry 4.0, devices atcontrol as well as supervisor level arefrequently connected to the Internet tocollect analytic data. The resulting net-work is called the “Internet of Things”(IoT) and “Industrial Internet of Things”(IIoT). Attackers seek to compromisesuch interconnected devices with mal-ware campaigns [1, 2] to use them for

spam distribution, Distributed Denial ofService (DDoS) attacks, cryptomining,or as an attack vector in AdvancedPersistent Threat (APT) attacks. For thisreason, interconnected devices areexposed to continuous threats andongoing attacks. The large set of diversehardware and software combined withthe neglection of security best practices,such as the use of the same default cre-dentials on all devices, the often non-existent update policies, and the lack ofsoftware hardening techniques renderIoT and IIoT devices an ideal target forattackers. Many solutions have beenproposed to monitor the Internet formalware infections. “Honeypots” are acommon practice but owing to the het-erogeneity of the devices they are sub-

stantially harder to implement in theIoT and IIoT domain than in the field ofcommodity systems (e.g., desktop com-puters, smartphones). The heteroge-neous landscape of IoT and IIoTdevices poses new challenges to thedeployment of honeypots.

The goal of the research projectAutoHoney(I)IoT [L1] is to provide aframework that automatically createstarget device tailored honeypots for the(Industrial) Internet of Things, whichare capable of convincing adversariesthat they are breaching a real deviceinstead of a decoy. The honeypots willbe executed in an emulation environ-ment that is able to interact with the out-side world over common IoT and IIoT

Autohoney(I)IoT - Automated device Independent

honeypot Generation of IoT and Industrial IoT devices

by Christian Kudera, Georg Merzdovnik and Edgar Weippl (SBA Research)

The heterogeneous landscape of IoT devices poses new challenges to the deployment of honeypots. So far no

generic honeypot framework exists that is capable of attracting attacks for the wide variety of hardware and

software architectures. By combining real world device information and virtualisation techniques, we aim to build

AutoHoney(I)IoT, a framework that automatically creates target device tailored honeypots for the (Industrial) Internet

of Things, which are capable of convincing attackers that they are breaching a real device instead of a decoy.

communication channels and allowsfine-grained supervision techniques tobe applied to monitor an adversary’sbehaviour throughout his entire attack.Figure 1 illustrates the overall conceptof the AutoHoney(I)IoT framework. Asinput, the framework requires thefirmware of the device to be virtualisedas well as a database containing realworld device information of commonmicrocontrollers and system on a chips(SoCs). The framework’s intendedfunctionality is to analyse a firmwaredump and find a Qemu [3] appliancecapable of executing the firmware.Furthermore, external peripherals (e.g.,network access) are attached as needed.

The analysis is based on a two-stepapproach. The aim of the first step is todetect the architecture (e.g., ARM,MIPS, PowerPC, x86) of the firmware.The main idea of the second step is toapply fine-grained supervision tech-niques during the execution to identifythe most appropriate processor.Therefore, the firmware is executedfirst on a generic processor correspon-ding to the identified architecture. Mostlikely this will result in an instructionmismatch, memory mismatch orinternal peripheral mismatch, sincethere is a plethora of differentembedded processors from various ven-dors with a wide variety of technicalcharacteristics (e.g., central processingunit types, memory maps). During theemulation an algorithm attempts toidentify an admissible embeddedprocessor that is capable of executingthe firmware dump. For this purpose,

we create a database containing the realworld technical characteristics ofmicrocontrollers and SoCs from variousvendors. The execution and fine-grained processor selection is repeateduntil a suitable processor is identified.

Since a convincing (I)IoT honeypotneeds to expose authentic systembehaviour and communicate with theoutside world, the AutoHoney(I)IoTframework will provide: (i) a method toattach communication interfaces to theemulated embedded processor andbridge them to physical as well as simu-lated hardware, and (ii) a method toattach custom external peripheraldevice models to a communicationinterface whose behaviour can bescripted in a simple way.

The FFG funded project started in July2019 and is expected to conclude inDecember 2021. It is jointly realised bySBA Research, FH Technikum Wien,TU Wien and Trustworks GmbH, all sit-uated in Vienna, Austria.

Link:

[L1] https://kwz.me/hEt

References:

[1] K. Angrishi: “Turning Internet ofThings(IoT) into Internet ofVulnerabilities (IoV): IoT Botnets”,arXiv Prepr. ArXiv1702.03681,2017.

[2] A. Spognardi et al.: “Analysis ofDDoS-Capable IoT Malwares”, inProc. of INSERT, 2017.

[3] F. Bellard: “QEMU, a fast andportable dynamic translator”,USENIX Annual TechnicalConference, FREENIX Track. Vol.41. 2005.

Please contact:

Christian Kudera, Georg Merzdovnik,Edgar Weippl, SBA Research, [email protected], ,[email protected],[email protected]



Figure 1: Overview of the AutoHoney(I)IoT framework.

“Yogurt” is an object-oriented declara-tive programming language for theInternet of Things (IoT). It allows theend-user to program their entire IoTecosystem through one environment byleveraging the capabilities of Igor[L1], an architecture for unified access

to IoT [1]. The language offers highexpressiveness by incorporatingmechanisms from traditional program-ming paradigms. Furthermore, theunderlying programming modeladopts a metaphor close to the users’real-life experience, thus reducing the

learning effort required to adopt thelanguage.

Traditionally IoT devices are pro-grammed through a variety of high-level languages. This, however, requiresan in-depth knowledge of computer

yogurt: A Programming Language

for the Internet of Things (IoT)

by Ivan H. Gorbanov, Jack Jansen and Steven Pemberton (CWI)

As IoT moves from the hands of professionals and academics into those of the general consumers,

it becomes increasingly important to provide them with the appropriate tools for interaction. Yogurt

is a domain-specific programming language for IoT, designed to tackle the disparity between

powerful but complex languages and user-friendly environments with restricted capabilities.


programming which often takes yearsof training and practice to develop. Inaddition, the decentralised nature of thedevelopment of these technologies hasled to the incorporation of many com-munication protocols and formal lan-guages. Therefore, managing a com-plete system often requires knowledgeof multiple programming environ-ments. Both the consumer industry andacademia have tried to address thisproblem by introducing systems whichact as a central control point for the cus-tomer’s IoT systems. Through these allof the user ’s devices can be pro-grammed through one language. Theseprogramming facilities, however, aremostly optimised for usability in orderto reduce the learning curve for inexpe-rienced users. This, therefore, leaves agap between highly expressive but diffi-cult programming languages and user-friendly environments, which lack suf-ficient expressive power. Consequently,the main purpose of this work istwofold: to extend the work of Jansenand Pemberton [1] by providing aviable programming language for Igor,and to fill the abovementioned gap thatcurrently exists in the programmingfacilities for IoT devices.

Yogurt’s underlying programmingmodel (Figure 1) achieves simplicity byproviding abstractions analogous to thereal world in order to reduce the gap

between what the programmer is tryingto achieve and how to achieve it. The“actor” abstraction represents alldevices which would make up an IoTsystem. Each “actor” has a state, whichrepresents what the device is doing inthe real world and can perform“actions” which change that state. An“action” gets triggered by a change inthe state of its own corresponding actoror that of another one. Conditions canalso be added in the form of “guards” toallow the device’s behaviour to varybased on its environment. This small setof general abstractions is at a highenough level so as to be easy to concep-tualise. Furthermore, they are generalenough to be applicable in the increas-ingly heterogenous collection ofdevices part of IoT.

The proposed model utilises severalmechanisms from established program-ming paradigms to allow it to tackle theuse cases that have come to be expectedby users of this technology. The actorabstraction works as a class in object-oriented programming in order to allowthe reuse of code, making writingYogurt programs more efficient.Furthermore, it allows for the use ofencapsulation and inheritance. Thismakes it easier to program more com-plex devices as a collection of simplerones. In addition, by forcing the pro-grammer to keep data and correspon-

ding methods together, it is easier tokeep track of dependencies and spot anyconflicts that may arise from differentmethods trying to change the same dataat the same time. Last but not least, thelanguage is declarative, meaning thatthe user needs to specify what the resultof the program needs to be rather thanhow exactly to achieve it. This featurebrings two major advantages: it furtherreduces the learning curve as the pro-grammer does not have to worry aboutconcepts such as memory managementand it makes the language context inde-pendent, meaning that a program whichswitches the lights on and off wouldremain the same regardless of the hard-ware devices used to implement thesolution as the result will be the same.

The proposed textual representation ofYogurt uses human readable keywordsto reduce the barrier to adoption andincrease code readability. Here is asimple example of a light which turnson with a presence detector (PD) onlywhen a day light sensor senses that it isdark outside:on(PD.present):

whenall(PD.present = True,

dayLight_sensor.night = True):

on <- True

The current version has been testedusing the Discount method for program-ming language evaluation [2].Participants confirmed that the lan-guage was easy to pick and use becauseit allows them to think about the task ina way in which they would in the phys-ical world, while suggesting changesthat could be made to increase effi-ciency.

Links:

[L1] https://github.com/cwi-dis/igor[L2] https://www.dis.cwi.nl/

References:

[1] J. Jansen, S. Pemberton: “Anarchitecture for unified access tothe internet of things”, XMLLONDON 2017 (2017).

[2] S. Kurtev, T. A. Christensen, B.Thomsen: “Discount method forprogramming languageevaluation”, in PLATEAU @SPLASH. 1-8, 2016.

Please contact:

Jack Jansen CWI, [email protected]

Figure 1: Yogurt programming model abstractions.

Figure 2: Light, presence sensor, daylight sensor example model.

TThe introduction of digital technolo-gies in economic and societal processesis key to addressing economic and soci-etal challenges such as ageing of popula-tion, ensuring societal cohesion, andsustainable development. While the fifthgeneration (5G) mobile communicationsare already upon us, the next steps intheir evolution will be key in supportingthis societal transformation, while alsoleading to a fourth industrial revolutionthat will impact multiple sectors.

IoT appears to be an important pillar of5G. Global networks like IoT createenormous potential for new generationsof IoT applications, by leveraging syner-gies arising through the convergence ofconsumer, business and industrialinternet, and creating open, global net-works connecting people, data, and“things”. A series of innovations acrossthe IoT landscape have converged tomake IoT products, platforms and

devices technically and economicallyfeasible. However, despite theseadvances, significant business and tech-nical hurdles must be overcome beforethe IoT’s potential can be realised.

Some important challenges and com-plexities include:

• Sustaining massively generated,ever-increasing, network traffic withheterogeneous requirements

• Adaptation of communication tech-nologies for resource-constrained vir-tualised environments

• Provision of networking infrastruc-tures featuring end-to-end connectiv-ity, security and resource self-config-uration

• Trusted information sharing betweentenants and host systems.

Overcoming these challenges requiresthe implementation and deployment

stack of IoT applications. The overallaim of SEMIoTICS [L1] is to develop apattern-driven framework [1-2], builtupon existing IoT platforms, to enableand guarantee secure and dependableactuation and semi-autonomic behav-iour in IoT/IIoT applications. TheSEMIoTICS framework supports cross-layer intelligent dynamic adaptation,including heterogeneous smart objects,networks and clouds. To address thecomplexity and scalability needs withinhorizontal and vertical domains,SEMIoTICS develops and integratessmart programmable networking andsemantic interoperability mechanisms.

The SEMIoTICS architectural frame-work (Figure 1) has been envisaged anddeveloped for efficient interconnec-tivity of smart objects. Each layer con-tains specific developed modules ableto handle different aspects and guar-antee different properties. More specifi-



Smart End-to-end Massive IoT Interoperability,

Connectivity and Security (SEMIoTICS)

by Nikolaos Petroulakis (FORTH), Konstantinos Fysarakis (Sphynx), Sotiris Ioannidis (FORTH), GeorgeSpanoudakis (Sphynx) and Vivek Kulkarni (Siemens)

Next generation networks, such as the Internet of Things (IoT), aim to create open and global networks

for connecting smart objects, network elements, applications, web services and end-users. Research

and industry attempt to integrate this evolving technology and the exponential growth of IoT by

overcoming significant hurdles such as dynamicity, scalability, heterogeneity and end-to-end security

and privacy. SEMIoTICS proposes the development of a pattern-driven framework, built upon existing

IoT platforms, to enable and guarantee secure and dependable actuation and semi-autonomic

behaviour in IoT/IIoT applications.

Figure 1: SEMIoTICS (i) envisaged architecture (ii) developed architecture.

cally, Software Defined Networking(SDN) Orchestration layer providesdata and control plane decouplingresulting in a cloud computing approachthat facilitates network managementand enables programmatically efficientnetwork configuration meeting dif-ferent IoT application requirementsrelated to security, bandwidth, latencyand energy efficiency, using semanticinformation. Network FunctionVirtualization (NFV) Orchestrationlayer provides a flexible, program-mable, dynamic and scalable net-working paradigm, making it ideal forsatisfying the QoS demands ofSEMIoTICS use cases. Field layer isresponsible for hosting all types of IoTdevices such as sensors and actuators aswell as IoT gateway which providescommon way for communication andensures enforcement of SPDI patternsin this layer. Finally, ApplicationOrchestration layer consists of all appli-cations receiving the communicationfrom field layer.

The above is validated by industry,using three diverse usage scenarios inthe areas of renewable energy, health-care, and smart sensing (Figure 2) andwill be offered through an open API. • Use Case 1 - Smart Energy: This use

case will showcase IIoT integrationin Wind Park Control Network pro-viding value added services such aslocal smart behaviour, predictive

maintenance and monitoring etc.Current state of the art of Wind Tur-bine Controller in a Wind Park con-trol network is typically an embeddedor highly integrated operating sys-tem, which rigorously follows devel-opment and pre-qualification prior todeployment in the real world.Because of this slow process, newfeatures, adding new sensors, actua-tors and related advancementsrequire several months or even yearsto be fully matured and operational inthe field.

• Use Case 2 – Smart Health Care: Thisuse case employs the SEMIoTICStechnologies to develop an Informa-tion and Communication Technology(ICT) solution aimed at sustainedindependence and preserved qualityof life for elders with mild cognitiveimpairment or mild Alzheimer’s dis-ease, with the overall goal of delay-ing institutionalisation: supportingboth “aging in place” (individualsremain in the home of choice as longas possible) and “community care”(long-term care for people who arementally ill, elderly, or disabled pro-vided within the community ratherthan in hospitals or institutions).

• Use Case 3 - Smart Sensing: This usecase offers an interesting specularapproach to this scenario (influencedby “Edge Computing” or “PervasiveComputing”). The main assumptionis that intelligent data processing

shall take place at sensor level, andthat distributed data classificationand clustering is a key aspect formassive system scalability. More-over, in this use case algorithmsderived from AI techniques will bedeployed at Gateway, down to MCUlevel, also allowing the system toonline/self-learn from the environ-ment. The latter is a quite challengingaspect in itself in the AI field.

SEMIoTICS is an IoT Security/PrivacyCluster European Union project fundedunder the H2020-IoT-03-2017 workprogramme, with Grant Agreementnumber 780315.

Link:

[L1]: http://www.semiotics-project.eu

References:

[1] N. E. Petroulakis, et al.:“SEMIoTICS ArchitecturalFramework: End-to-end Security,Connectivity and Interoperabilityfor Industrial IoT”, in 2019 IEEEGlobal IoT Summit (GIoTS), 2019.

[2] K. Fysarakis, et al.: “ArchitecturalPatterns for Secure IoTOrchestrations”, in 2019 IEEEGlobal IoT Summit (GIoTS), 2019.

Please contact:

Nikolaos Petroulakis, ICS-FORTH,Greece, [email protected]


Figure 2: SEMIoTICS Use Cases.

Today’s world is more connected thanever: both objects of production andeveryday objects are connected andexchange information. This raises manychallenges for wireless sensor networks(WSN) to ensure the proper quality ofservice under low power constraints, forexample when the sensors are mobile.Surveys have shown that two thirds ofapplications involve some sort ofmobility (e.g. cold chain management,environmental monitoring, vehicletracking) [1]. Compared with telephonyand internet, this problem has receivedrelatively little attention in the literature,and although it has been the subject ofongoing surveys and technologicalstudies like ours, no standard solutionsyet exist [2].

Due to their limited resources, sensorsare designed to communicate withoutany infrastructure and each sensor routespackets of his neighbour to reach theborder router, which is the gatewaybetween the WSN and the Internet.Modern sensors are smart and can com-municate directly with external services

resulting in a very large range of appli-cations from smart cities and manufac-turing to buildings and home automa-tion. Two types of mobility can be dis-tinguished. Micro mobility refers tosensor movement within the samesensor network domain with no changein the network address prefix. Macromobility refers to sensor movement to anew sensor network domain withanother prefix. Our current scope ofwork is micro mobility.

To exchange information in a WSN, theRPL protocol (Routing Protocol forLow-Power and Lossy Networks) canbe considered [L1]. Since it is designedfor lossy networks, it adapts the routingstrategies to any topological changesdynamically, meaning it providesimplicit mobility support. However, thismanagement is not efficient because itcan take a long time to update thetopology and reach the new location ofa mobile sensor. Our work aims toenhance the global mobility manage-ment in the context of RPL-based WSNand increasing the network resilience.

Our proposal is based on 6LBR(6Lowpan Border Router) referenceimplementation [3] provided in OpenSource [L2]. It allows multiple borderrouters to aggregate WSNs in order toincrease resilience. Among the avail-able modes, we consider the smartbridge (SmartBridge) mode where eachborder router (BR 1 to 3) manages itsWSN independently as depicted inFigure 1(a). When a mobile sensormoves from BR1 to BR2, BR2 updatesits network configuration and createsnew routes so the mobile sensor cancommunicate in this new location. AsBR1 is not aware of this update and thenode will stay in its routing table until atimer expiration is reached after severalattempts to join it. Our proposal imple-ments a synchronised smart bridge(SyncSmartBridge) mode based on thevirtual root principle combined to a syn-chronisation mechanism. The goal ofthe virtual root is to provide a unifiedcontrol plane view of all the concreteWSN within the domain. The synchro-nisation mechanism will ensure thebroadcast of synchronisation triggers to



Towards Improved Mobility Support in Wireless

Sensor Networks

by Amel Achour, Lotfi Guedria and Christophe Ponsard (CETIC)

Sensor networks are developing at a fast pace and are facing new challenges such as sensor

mobility resulting in topological changes within a network or in adjacent networks with roaming

nodes. To address such situations, we are working to extend a reference open source

implementation of a border router with extended smart bridge mode.

Figure 1: 6LBR based mobility management.

other BRs in order to update theirtopology and inform them about thenew location of the mobile sensor. Asshown in Figure 1(b), when a sensormoves from BR1 to BR2, it triggers anetwork configuration update in BR2.At the same time, BR2 broadcasts asynchronisation message to the otherBRs so their topology and the sensorlocation is immediately updated.

Our protocol design is being validatedusing the Cooja simulator [L3]. Thecurrent work is focused on the scala-bility, performance assessment androbustness tests through a demonstratorthat covers various baseline scenarios:one or multiple sensors moving in

same/different directions, simultane-ously or randomly. We also plan to con-sider a combination of more complexmobility scenarios to measure synchro-nisation delays in situations of highmobility and high network density.

Links:

[L1] https://kwz.me/hES[L2] https://github.com/cetic/6lbr [L3] https://anrg.usc.edu/contiki/

References:

[1] J. Wang, et al.: “A survey aboutrouting protocols with mobile sinkfor wireless sensor network”, Int. J.of Future Generation Communica-tion and Networking, 2014;7(5).

[2] A. Achour, L. Deru, J-C. Deprez:“Mobility management for wirelesssensor networks a state-of-the-art”,Procedia Computer Science 52(2015).

[3] L. Deru, et al.: “Redundant BorderRouters for Mission-Critical6LoWPAN Networks”, Proc. of theFifth Workshop on Real-WorldWireless Sensor Networks, 2013.

Please contact:

Lofti Guedria, CETIC, Belgium+32 472 56 62 [email protected]


Processing telemetry data is a challengein any IoT architecture [1]. In the spacesector, monitoring remote sensors is acommon problem. The recent increasein downlink bandwidth and availableprocessing power on spacecrafts haveresulted in an increase in the number andthe volume of telemetry measurementsavailable to spacecraft controllers tomonitor the health and safety of a space-craft. In practice, the number of mainte-nance parameters is generally in the tensof thousands range. While providing awealth of diagnostic information, thesehuge numbers overwhelm the ability ofthe human brain to oversee all telemetrymeasurements. Automatic checks andcomputer-aided data analysis can pro-vide a daily overview for spacecraftoperators and operations engineers.

In order to efficiently operate satelliteconstellations as well as spacecrafts, theLuxembourg Institute of Science andTechnology [L1] and the Institute ofAstronomy of KU Leuven [L2] havedeveloped the “Decision ManagementSystem for Safer Spacecrafts” (DMSS)platform for the European SpaceOperations Centre (ESOC), the mainmission control centre of the European

Space Agency (ESA). DMSS offers aself-learning visual platform foranomaly detection in telemetry datacoming from embedded sensors.

Based on complex data analytics algo-rithms [2], DMSS follows a visual ana-lytics approach [3], providing interac-tive and visual representations oftelemetry data and derived computa-tional data. These representations comein the form of diagrams, such asPoincaré plots, KDE schemas, Heatmaps, and time series plots (Figure 1),which are interactive and connected:choosing data in one diagram highlightscorresponding data in another relateddiagram.

To help monitor the sensors, the mainscreen of DMSS presents a heat mapdisplaying all telemetry anomaly scoresfor a specific day (Figure 2). The goal isto quickly identify which telemetryparameters potentially show an unusualbehaviour:• Blue means that no anomaly score

was calculated for this date (due tomissing or incomplete data – which iscommon with telemetry data fromremote sensors).

• Grey means that the anomaly score iszero, thus the corresponding parame-ter has a standard behaviour for theselected date.

• Red means that the anomaly score ispositive, this parameter has anuncommon behaviour; the usershould click on it to see more detailsand investigate whether it is due to areal anomaly.

• Green means that the anomaly scoreis negative. Negative scores mayoccur for some score types when theanomaly score was previously posi-tive and now the scoring systemdetects that the parameter behaviouris coming back to normal.

With this view, end-users can see if agroup of telemetry parameters has ahigh anomaly score and see if one spe-cific telemetry parameter has a highscore compared to other similartelemetry parameters.

As a use-case, the data from two spacemissions operated by the EuropeanSpace Agency were analysed: MarsExpress (5,127 telemetry sensors i.e.

designing IoT Architectures: Learning from

Massive Spacecraft Telemetry data Analytics

by Olivier Parisot, Philippe Pinheiro and Patrik Hitzelberger (Luxembourg Institute of Science and Technology)

“Decision Management System for Safer Spacecrafts” (DMSS) is a data analytics platform for the space

domain that can detect anomalies in huge telemetry data acquired from numerous sensors in a complex

IoT architecture.

141 GB of data) and GAIA (28,209telemetry sensors i.e. 1.46 TB of datafor GAIA - including raw time seriesand statistical pre-calculations). Forexample, the number of points per timeseries is very variable, from a few sam-ples up to 13 million samples. DMSSwas able to display the potentiallydetected anomalies for the MarsExpress mission. It also helped expertusers to visually explore this importantamount of data in a new way and try tofind correlations between a potentialdetected anomaly and other parameters

like events or commands sent to thespacecraft.

To conclude, DMSS helps to analyselarge volumes of time series datacoming from spacecraft sensors - whichis a typical scenario for IoT applica-tions. In this use case, operational risksare high, and efficient support of space-craft operator engineers is of paramountimportance. It is therefore crucial thatsoftware architecture and semi-auto-matic data analytics support performefficiently and reliably. We anticipate

that the approach could be applied to alarge extent in other IoT scenarios.

Links:

[L1]https://www.list.lu/en/cooperations/lines-of-business/space/[L2] https://fys.kuleuven.be/sterReferences:

[1] Lazarescu et al.: “Design of aWSN Platform for Long-TermEnvironmental Monitoring for IoTApplications”,https://doi.org/10.1109/JETCAS.2013.2243032

[2] Royer et al.: “Data miningspacecraft telemetry: towardsgeneric solutions to automatichealth monitoring and statuscharacterisation”,https://doi.org/10.1117/12.2231934

[3] Keim et al.: “Visual analytics:Definition, process, andchallenges”,https://doi.org/10.1007/978-3-540-70956-5_7

Please contact:

Olivier Parisot, Philippe Pinheiro,Patrik Hitzelberger Luxembourg Institute of Science andTechnology [email protected], [email protected], [email protected]



Figure 1: DMSS provides a

dashboard to show telemetry

data coming from remote

sensors.

Figure 2: The heat map helps to the potential anomalies in telemetry data coming from these

sensors.


The French company Sencrop [L1] aimsto help farmers in their hard daily work[1]. Sencrop manufactures and sellsweather stations that autonomously col-lect data to help farmers accurately fore-cast the weather (e.g. risk of frost) andmake sound decisions for the crop cycle(requirements for watering, fertiliser,etc). These stations collect accurateparcel-specific weather-related data,such as temperature, pluviometry,humidity, wind speed, etc. They rely onbatteries that cannot be easily rechargedor replaced regularly. Therefore, theSigfox wireless technology is used tosend the sensed data to the internet [L2].Sigfox technology enables long rangecommunications at a low energy expen-diture. The Sigfox base stations to whichthe weather stations offload their dataare deployed around the world to pro-vide connectivity to customers.

But these advantages do not comewithout a price. Sigfox operates on theunlicensed ISM radio bands, whichlimits the time that a radio transceivercan emit by law. Sigfox technology thussuffers from low data rates and asym-metric wireless links. Furthermore, theSigfox base stations are not yet deployedworldwide.

Sencrop weather stations require aworldwide ubiquitous network access toassist any farmer anywhere on Earth, butthe coverage provided by Sigfox isinsufficient. Furthermore, Sencropneeds to remotely upgrade the devices’firmware once a station is deployed;Sigfox can handle a small volume ofregular monitoring data but is notequipped to support a whole firmware.

Indeed, the main issue faced by Sencropis how to extend the current coverageand enable a bi-directional connectivityto the weather stations whilst limitingthe energy consumption of devices. Onepossible solution is a weather stationthat integrates several wireless commu-

nication technologies, each providingdifferent characteristics (delay, datarate, energy consumption, etc.),enabling the station to use the mostappropriate technology for any giventype of data to be sent.

As depicted in Figure 1, weather station“A” can reach Sigfox, as well as alter-native technologies T1 and T2, and canautonomously pick the one that bestsuits its data requirements. T1 offers alarge data rate and a bidirectional con-

nectivity but with a high energy expen-diture. It will be used for exceptionallylarge amounts of data, such as afirmware upgrades. T2 will only beused to send urgent data such as analarm because it provides low delays atthe cost of a medium energy consump-tion. Finally, Sigfox that features longdelays and small data rates but benefitsfrom a low power expense will befavoured for sending frequent and reg-ular monitoring data. This solves theaforementioned issues, as station B

Autonomous Collaborative Wireless Weather

Stations: A helping hand for Farmers

by Brandon Foubert and Nathalie Mitton (Inria)

In the context of smart farming, communications still pose a key challenge. Ubiquitous access to the

internet is not available worldwide, and battery capacity is still a limitation. Inria and the Sencrop

company are collaborating to develop an innovative solution for wireless weather stations, based on multi-

technology communications, to enable smart weather stations deployment everywhere around the globe.

B

???

Sigfox

T1

T2

C

A

Figure 1: Autonomous multi-technology stations. A, B and C are weather stations. T1, T2 and

Sigfox are different wireless techonologies.

Figure 2: Multi-hop multi-technology network. D, E and F are weather stations. T1, T2 and

Sigfox are different wireless technologies.

which is not in range of the Sigfox net-work can still use the other two tech-nologies.

But this solution raises other issues.Adding wireless technologies and theability to switch from one to anotherconsumes energy and processing capa-bilities, resources that are limited onSencrop stations. Moreover, stations arestill dependent on a direct link to a basestation. Areas deprived of any communi-cation capability still exist on our planetin isolated rural environments. Inria andSencrop jointly address this excitingchallenge by designing a joint smart net-work interface selection, and a multi-hop multi-technology routing mecha-nism. It will dynamically adapt the dataforwarding according to the local envi-ronment (remaining energy, connec-tivity, etc.) and the importance of thedata by respecting environmental andhardware limitations of each station.

As depicted in Figure 1, if a station isisolated, such as station C, it may stillbe in the vicinity of another station, likeB, through which it can reach a base sta-tion. This mechanism will not onlyextend the network coverage, but alsosave power and alleviate networktraffic. Indeed, instead of using energyintensive long communication links, adata packet can follow a suite of moreenergy efficient smaller ones. Forinstance in Figure 2, station D canchoose to offload its data through E andF (technology T2) and then to the basestation (Sigfox) which might be lessenergy costly than using the direct linkT1 to the base station. Stations will notonly have to choose which wirelesstechnology to use autonomously, butalso to which neighbour they send theirdata to satisfy the data requirements.

This system is expected to improve theoverall energy efficiency of Sencrop

stations and enable a broader deploy-ment to assist farmers with smart con-nected things everywhere around theglobe.

This work was partially supported by agrant from CPER DATA, Sencrop,FEDER and I-SITE.

Links:

[L1] https://sencrop.com/en/[L2] https://www.sigfox.com/en

Reference:

[1] A. Gregoire: “The mental health offarmers”, Occupational Medicine, Vol.52, Issue 8, 2002, p. 471–476,https://doi.org/10.1093/occmed/52.8.471

Please contact:

Brandon FoubertInria Lille, France+ 33 3 59 57 79 [email protected]



The distinctive speech characteristics ofpeople suffering from allergic rhinitishave the potential to enable remote diag-nosis using speech analysis technology.We have developed a system for thispurpose, consisting of two experts: thefirst provides a robust estimate of thejitter of the fundamental frequency andthe second provides a confidence scorederived from Hidden Markov Model-based acoustic modeling. These expertsare calibrated using two sets of speechdata, one from patients with allergiesand a second from the same individualsafter treatment. This supervised calibra-tion allows the parameters of the twoexperts’ algorithms to be fine-tuned andenables some thresholds to be derived,which are subsequently used to diag-nosis others.

The notion of the rank of a signal, that isneeded in this work, is defined bymeans of the rank of its covariancematrix [1,2]. Given the signal s and the

nth order covariance matrix Cn, therank of the signal s is equal to p, wherep is the rank of the covariance matrixCn and p < n. In case p = n for all n, thesignal s is said to be full rank. Full ranksignals are more complex to handle andare converted to limited rank by meansof a singular value decomposition(SVD) based successive projectionsalgorithm [3].

The fundamental frequency estimationalgorithm is highly accurate and con-sists of two steps. During the first stepthe goal is to obtain from the originalspeech signal a new one where the fun-damental frequency is the predominantfrequency. This signal enters the secondstep of the algorithm that provides anapproximate rank two signal of the fun-damental frequency.

The first step is based on an iterativezero-phase filtering whose frequencyresponse is a monotonically decreasing

function. This guarantees that only theenergy around the fundamental fre-quency remains as the number of itera-tions increases.

For the second step the SVD algorithmto reduce the rank of a signal by succes-sive projections is used. Finally, thejitter of the fundamental frequency iscomputed as the variance of the dis-tances between consecutive peaks of therank two signal (Figure 1).

The second expert uses acoustic modelsfor the derivation of confidence scoresfor the uttered phrases. These modelsprovide estimates of the probability offeatures extracted from speech and givea string of phonemes. Each phoneme isrepresented by a Hidden MarkovModel. For the Greek language, 32phonemes are used to describe the var-ious pronunciations. The acoustic mod-elling not only provides the phonetictranscription of the utterance but also

Automatic detection of Allergic Rhinitis

by Gregory Stainhaouer, Stelios Bakamidis and Ioannis Dologlou (RC ATHENA)

The spectral characteristics of speech can be used to cluster individuals according to whether or not

they suffer from an allergy. Based on the principles of adaptive modelling and fundamental frequency

variations, as well as speech analysis by means of acoustic models, our technique achieves an

efficient classification based on uttered speech over a mobile phone. The final decision is derived by

combining the individual estimates, providing a tool for the automatic diagnosis of allergies.


the probability of each phoneme whichreflexes the confidence score.

The automatic diagnosis system wasfirst trained and then tested using reallife data collected from patients. For thetraining phase sixteen patients wereused, who returned at a later stage whenthey were cured and uttered the samephrases. The two sets of experts thatneed to be trained are the jitter and theconfidence score of the acousticmodels.

In order to handle new patients, themean values of the jitter for healthy andunhealthy subjects are computed,denoted by mjc and mjs respectively.

These values are mjs=16 and mjc = 12.If jn stands for the jitter of the newpatient then if jn > mjs the patient isunhealthy, whereas if jn < mjc thepatient has no problem. For values of jnbetween mjs and mjc no accurate deci-sion can be made (Figure 2).

The phrases are also processed by theacoustic models and a mean confidencescore is produced. By denoting mps, themean value that a patient gets beforetreatment, and mpc, the mean value thatthe same patient gets after treatment,one expects to find mpc > mps. A globalmean value for all cured patients mpcgand a global mean for all unhealthypatients mpsg is computed to enable the

classification of a new patient accordingto the confidence score criterion. Thesevalues are mpsg = 17.7 and mpcg =18.7. If mpn stands for the mean confi-dence score of the new patient then ifmpn < mpsg the patient is unhealthy,whereas if mpn > mpcg the patient hasno allergy. For values of mpn betweenmpcg and mpsg no accurate decisioncan be made (Figure 3). A simple linearoptimisation technique is used to com-bine the two experts in order to providean overall decision.

The algorithms were developed in theResearch Center ATHENA in AthensGreece. The work started in October 2018and the first version was delivered in June2019. The project “Patient AllergyTracer” (project code: Τ1ΕΔΚ-02436)supported this work, which is imple-mented under the Action “Research,Create, Innovate”, funded by theOperational Program “Competitiveness,Entrepreneurship and Innovation” (NSRF2014-2020) and co-financed by Greeceand the European Union (EuropeanRegional Development Fund).

Future activitiesFuture plans focus on the performanceof the algorithms both in terms of accu-racy and speed. Improving the accuracyinvolves a better training process byincluding more patients in the experi-ment and also ameliorates the perform-ance of the acoustic models. To improvethe speed, fast SVD algorithms areneeded for the implementation of thesuccessive projections algorithm.

References:

[1] I. Dologlou, G. Carayannis:“Physical interpretation of signalreconstruction from reduced rankmatrices”, IEEE ASSP, July 1991,pp. 1681-1682.

[2] I. Dologlou, S. Bakamidis, G.Carayannis: “Signal decompositionin terms of non-orthogonalsinusoidal bases”, SignalProcessing, Vol. 51, June 1996.

[3] J.A. Cadzow: “Signalenhancement: A compositeproperty mapping algorithm”,IEEE, ASSP, Vol. 36, No. 1,January 1998, 49-62.

Please contact:

Ioannis DologlouRC ATHENA, Greece+302106875306, [email protected]

Figure 1: The top figure depicts a

frame of voiced speech and the

bottom figure depicts the

corresponding rank two signal.

Note that peaks are very clear

and distances between peaks can

easily be computed to ensure an

accurate value for jitter.

Figure 2: The solid line depicts

mjc, the mean of healthy

condition and the dashed line

depicts mjs, the mean of allergic

condition. The various + symbols

depict unhealthy patients. Most of

them lie beyond mjs and are

safely classified. Many lie

between the two means and

decision becomes ambiguous and

few are misclassified below mjc.

Figure 3: The dashed line depicts

mpcg, the mean of healthy condi-

tion and the solid line depicts

mpsg, the mean of allergic condi-

tion. The various + symbols de-

pict unhealthy patients. Most of

them lie below mpsg and are

safely classified. Some lie be-

tween the two means and deci-

sion becomes ambiguous. There

are also some misclassified cases

lying above mpcg.

The use of health and well-being tech-nologies in smart IoT ecosystems hasbeen steadily increasing, and they arenow found in environments such assmart homes, age-friendly workplacesand public spaces [1]. The deploymentof smart sensors aims to support func-tional, physiological and behavioralmonitoring, which can benefit botholder adults facing gradual degradationof their motor and cognitive skills due toaging and workers in harsh environ-ments performing arduous, stressful orhealth hazardous tasks [2]. Such smartsystems can promote a reactive livingand working environment, which pro-vides appropriate and timely recommen-dations, acts preventively and mitigateshealth risks.

sustAGE [L1] is a H2020 EU projectthat aims to develop a person-centeredsmart solution fostering the concept of“sustainable work” for EU industries,thus supporting the well-being, well-ness at work and productivity of ageingemployees along three main dimen-sions. The first dimension is directedtowards improving occupational safetyand health based on workplace andperson-centered health surveillancemonitoring. The second aims to pro-mote the well-being of employees viapersonalised recommendations forphysical and mental health whilst thethird supports decision making relatedto task/job role modifications aiming tooptimise overall workforce produc-tivity. The sustAGE solution is

deployed in two challenging industrydomains, specifically: (i) manufac-turing, focusing on car assembly lineworkers; and (ii) transportation & logis-tics, focusing on dock workers involvedin vessel loading/unloading operations.

The system functionalities build uponan IoT ecosystem, based on off-the-shelf sensors integrated into dailydevices and in the work environment,considering both indoor (manufac-turing) and outdoor (port) working con-ditions. The system gathers contextualinformation from the working environ-ment and users’ physiological signals,tasks, activities and behavioural pat-terns, in order to support user profilingand provide personalised recommenda-



Supporting the Wellness at Work and Productivity

of Ageing Employees in Industrial Environments:

The sustAGE Project

by Maria Pateraki (FORTH-ICS), Manolis Lourakis (FORTH-ICS), Leonidas Kallipolitis (AEGIS IT Research),,Frank Werner (Software AG), Petros Patias (AUTH) and Christos Pikridas (AUTH)

Industrial environments can benefit from smart solutions developed on top of an infrastructure

combining IoT and smart sensors that monitor workers in an non-invasive way, allowing the early

detection and prevention of health risks. The sustAGE project aims to improve occupational safety and

workforce productivity through personalised recommendations in two key industrial environments.

Figure 1: sustAGE system architecture.

tions. Measurements collected from dif-ferent devices and system modules sup-port the definition of key micro-moments related to the user’s dailyschedule, work environment, workload,physical/emotional/mental state andsocial activities, whilst taking intoaccount associated temporal variables.The IoT infrastructure comprises of: (i)Environmental sensors measuring airtemperature, humidity, air quality, pres-sure, dust concentration and noise withcustom, low cost Raspberry Pi/Arduinosensors; (ii) Passive cameras installed inkey working locations, specificallystereo cameras for monitoring postureand repetitive user actions in the indoormanufacturing environment and monocameras in the outdoor port environ-ment for monitoring crane operatorsand dock workers involved in vesselloading/unloading; (iii) Beacons forlocalisation in indoor environments,achieving a precision of up to 10-20 cmwithin a range of 100 m; (iv) GNSSreceivers built in smartphones for local-isation in outdoor environments; (v)Wristwatch devices gathering physio-logical measurements, able to delivernotifications to users from the system;and (vi) Galileo-enabled Smartphonedevices, offering centimetre accuracyand ability to communicate with thewristwatches.

The aforementioned devices/sensorscollaboratively provide information ondifferent user activities/actions (e.g.walking, bending, standing/sitting,pushing/pulling objects) and states (e.g.fatigue, discomfort), integrating tem-poral aspects and detecting specificevents in the environment (e.g. userpresence in specific areas, proximity tohazardous conditions). Moreover, thesmartphone is the primary device forcommunication and multimodal inter-action supporting natural languageunderstanding and voice sentimentanalysis. The adopted IoT configurationexhibits the advantages of unobtrusiveuser context interaction monitoring in aprivacy-preserving way, since in privatelife, outside the working environment,only the wristwatch and the smartphoneare used.

Figure 1 shows the system’s architec-ture, which is conceptually divided intofour layers, namely monitoring,streaming, personalisation and recom-mendation. The monitoring layerincludes components that receive raw

data from various sources, supportingprocessing near the end-devices thatprevents potentially privacy-sensitiveinformation from being sent to thesystem’s upper layers. Cameras, wrist-watches, environmental and locationsensors as well as users’ speech com-prise the list of data sources that feedthe system. The streaming layerincludes the sustAGE “Bridge andUniversal Messaging” modules thatingest the data through a securegateway and prepare them for distribu-tion to the rest of the system. Datastreams of different protocols arebuffered, homogenised and further sentto subsequent components via acommon communication protocol thatensures speed and scalability in deliveryof near real-time data. The UniversalMessaging bus supports streaming ofthe real-time data, through thetopics/messages across the IoT infra-structure and all other platform compo-nents. The personalisation layerincludes personalisation mechanismsgenerating new information, throughdistilling the short-, medium-term statesand long term traits, extracting knowl-edge abstractions and enabling user pro-filing to be updated regularly byexploring association with pastepisodes. The Apama streaming ana-lytics engine [L2] performs analytics onthe incoming data, referencing histor-ical information where necessary, toidentify previously occurred patterns.Generated knowledge is stored in therespective knowledge base to con-stantly improve analytics, reasoningand thus user recommendations. Therecommendation layer consists of rea-soning and recommendation moduleswhich determine user recommenda-tions, taking into account spatiotem-poral constraints.

Environmental sensor measurementsalong with user state parameters are col-lected by the IoT platform of thestreaming layer through dedicated gate-ways for monitoring and analytics. InsustAGE, the main requirements of theIoT management solution relate to: (i)interoperability, that is the ability toacquire data from devices from dif-ferent vendors that use different proto-cols; (ii) scalability, i.e. the ability tohandle increased amounts of data; and(iii) security and privacy of the commu-nicated data. Apart from these chal-lenges, efforts are focused on robustdata analysis techniques to efficiently

handle the data streams and support thedetermination of proper actions.

sustAGE is a multidisciplinary projectinvolving ten European partners:Foundation for Research andTechnology – Hellas (Greece, coordi-nator), Centro Ricerche Fiat, (Italy),Heraklion Port Authority (Greece),Software AG (Germany), ImaginarySrl. (Italy), AEGIS IT Research UG(Germany), Leibniz Research Centrefor Working Environment and HumanFactors (Germany), University ofAugsburg (Germany), NationalDistance Education University (Spain)and Aristotle University of Thessaloniki(Greece).

Links:

[L1] https://www.sustage.eu[L2]https://en.wikipedia.org/wiki/Apama_(software)

References:

[1] A. Kumar, H. Kim and G.P.Hancke: “Environmentalmonitoring systems: A review”,IEEE Sensors Journal 13(4):1329-1339. 2013.https://www.doi.org/10.1109/JSEN.2012.2233469

[2] A. Almeida, et al: “A criticalanalysis of an IoT—aware AALsystem for elderly monitoring”,Future Generation ComputerSystems, Vol.97: 598-619, 2019.https://www.doi.org/10.1016/j.future.2019.03.019.

Please contact:

Maria Pateraki, Manolis Lourakis FORTH-ICS, [email protected]@ics.forth.gr

Leonidas KallipolitisAEGIS IT Research, [email protected]

Frank Werner Software AG, [email protected]

Petros Patias, Christos PikridasAristotle University of Thessaloniki(AUTH), [email protected], [email protected]


Research and Innovation


Eu

rop

ea

n R

ese

arc

h a

nd

In

no

va

tio

nhigh Performance

Software defined Storage

for the Cloud

by Thomas Lorünser (AIT), Stephan Krenn (AIT) andRoland Kammerer (Linbit HA-Solutions GmbH)

Distributed Replicated Block Device (DRBD) is the de

facto standard for redundant block storage. It is used

more than 250,000 times world-wide and part of the

official Linux kernel. DRBD4Cloud is a research project

which aims at increasing the applicability and

functionality of DRBD in order to enter new markets and

to face future challenges in distributed storage.

Redundant data storage is a necessity for business continuityof virtually any cloud service. An intuitive approach is fulldata replication to multiple storage nodes, which is currentlydone by DRBD [L2]. However, due to large bandwidthrequirements and storage overhead this is not feasible forlarge-scale deployments with many mirroring nodes, i.e.,typical cloud settings. Other than DRBD, Ceph is the majorbackend block storage implementation. Ceph already offersintegration with OpenStack. However, Ceph’s performancecharacteristics prohibit its deployment in certain low-latencyuse cases, e.g., as backend for Oracle MySQL databases.

DRBD4Cloud [L1] will increase the performance and scala-bility (technical and organisational) of highly-available soft-ware defined block storage in dynamic large scale clouddeployments. It is based on DRBD, which is a high-perform-ance low-latency low-level building block for block replica-tion, offering key functionality of such systems. During theproject DRBD will be integrated into OpenStack and DC/OS(the Distributed Cloud Operating System), the most preva-lent tool suites to manage distributed computing resources.This will enable cloud backend providers to use DRBD tech-nologies for replicated block storage, which is an inherentlyneeded building block for highly reliable cloud storage offer-ings.

To ease the deployment and maintenance of DRBD a collec-tion of key components will be offered as an easy-to-usesoftware package. A component to orchestrate and monitorthe storage environment consisting of a multitude of DRBDnodes will be developed. On top of this, web-based and API-based management and monitoring solutions will be devel-oped in order to ease adoption of DRBD-based software-defined storage for cloud environments.

To guarantee availability, DRBD is currently storing up to 32full data replicas on remote storage nodes. DRBD4Cloudwill allow for the usage of erasure coding, which allows oneto split data into a number of fragments (e.g., nine), such thatonly a subset (e.g., three) is needed to read the data. This willsignificantly reduce the required storage and upstream band-width (e.g., by 67 %), which is important, for instance, forgeo-replication with high network latency. Additionally, spe-cific schemes called secret sharing can even guarantee thatthe servers do not learn anything about the plain data,


without requiring cryptographic keys [1]. This will allow forusing public cloud storages without compromising confiden-tiality, making DRBD also usable to SMEs without own datacentres but requiring highly available storage.

The feasibility of cloud integration of DRBD for smallsetups has already been demonstrated by means of a firstproof-of-concept prototype which showed that the mainchallenge is to meet the scalability requirements of largescale deployments. Another challenge is the design andmaintenance of a common shared driver core between theOpenStack and DC/OS integrations. Additionally, con-cerning erasure coding and secret sharing the main challengeis minimising the impact upon latency. Fortunately, in suchschemes, read operations only require access to a subset ofthe storage nodes in order to retrieve the data, which posi-tively affects the availability and latency. A first result onhardware acceleration of secret sharing [2] also shows thepotential of low latency implementations on modern high-bandwidth network interface cards. As an optimisation, themonitoring solution could be utilised to optimise the globalworkload of the overall storage cluster and as results in [3]show, efficient auditing mechanisms can be used to verifythe data integrity in the system although erasure coding andsecret sharing are applied.

In summary, with DRBD4Cloud a new and highly optimisedout-of-the-box solution for multi-node software definedstorage in high-load cloud environments will be developed.The results will be delivered as software implementations,which similar to DRBD will (mostly) be published under anopen source license. The extensions will allow DRBD – thede-facto standard for distributed replicated block devices – tocut the storage overhead by 50-80 % while guaranteeingpractically equivalent levels of redundancy and will allowconfidential data to be securely stored on semi-trusted cloudproviders. Simplifying the deployment and enabling moni-toring as well as integrating DRBD into cloud platforms willallow cloud providers to pick up the technology. At the endof DRBD4Cloud, all extensions will be delivered as inter-nally tested prototypes. DRBD4Cloud is a joint effort of

Linbit HA-Solutions GmbH, Pro-zeta A.s. and AIT AustrianInstitute of Technology (AIT) which has received fundingfrom the EUREKA Eurostars programme [L3].

Links:

[L1] https://kwz.me/hE3[L2] https://www.linbit.com [L3] https://www.eurostars-eureka.eu/project/id/11450

References:

[1] T. Lorünser, A. Happe, and D. Slamanig: “ARCHIS-TAR: Towards Secure and Robust Cloud Based DataSharing”, in CloudCom 2015, IEEE,https://doi.org/10.1109/CloudCom.2015.71

[2] J. Stangl, T. Lorünser, S.M. Dinakarrao: “A fast andresource efficient FPGA implementation of secret shar-ing for storage applications”, DATE 2018, pp.654–659, https://doi.org/10.23919/DATE.2018.8342091

[3] D. Demirel, et al.: “Efficient and Privacy PreservingThird Party Auditing for a Distributed Storage System”,ARES 2016, pp. 88–97,https://doi.org/10.1109/ARES.2016.88

Please contact:

Stephan KrennAIT Austrian Institute of Technology [email protected]

Figure 1: Depending on the configuration, no individual share contains any sensitive information about the replicated data,

while the data can be recovered from any two of the shares to achieve high availability.

Research and Innovation

The Battle of the video

Codecs in healthcare

by Andreas S. Panayides (Sigint Solutions and Universityof Cyprus), Marios S. Pattichis (University of New Mexico)and Constantinos S. Pattichis (University of Cyprus)

Video compression is the core technology in mobile

(mHealth) and electronic (eHealth) health video

streaming applications. With global video traffic

projected to reach 82 % of all internet traffic by 2022,

both industry and academia are struggling to develop

efficient compression algorithms to match these

unprecedented needs. To the best of our knowledge, this

is the first performance comparison of emerging VVC

and AV1 video codecs for medical applications.

Video codecs are currently experiencing unparalleled levelsof growth driven by unprecedented video traffic demandsthat are projected to reach 82 % of all internet traffic by 2022[1]. This growth is facilitated by advances in open-sourcevideo delivery protocols, such as Dynamic AdaptiveStreaming over HTTP (MPEG-DASH) and WebRTC (Real-Time Communication (RTC)). Most importantly, industryinitiatives such as Google’s WebM project (leading VP8/ 9video codecs development), and subsequent formation of theAlliance for Open Media (AOM) [L1] comprising keyindustry affiliates, have created a highly competitive envi-ronment, investing efforts towards creating the first ever roy-alty free video codec. In this context, AOM announced thecode freeze of its debut encoder, AV1, in 2018, claiming thebest encoding performance to date. At the same time, theJoint Video Experts Team (JVET) formed in October 2017that has undertaken the development of H.265’s successor,termed Versatile Video Coding (VVC), has just released itsreference software abbreviated VTM (VVC Test Model)[L2].

In healthcare, video compression is a key enabling tech-nology that is widely used for real-time medical video com-munications in a mobile-health setting [2], [3]. Applicationscenarios range from remote diagnosis and care to emer-gency incidence response, medical education, tele-robotics,and second opinion provision. A plethora of applications fur-ther extends to the home setting for assisted living applica-tions. Moreover, video compression is also key in low-band-width applications such as remote diagnosis from isolatedlocations, developing countries, and disaster sites.

The objective of this study was to investigate the compres-sion efficiency of well established (VP9 and H.265 (alsotermed high efficiency video coding (HEVC)), recently stan-dardised (AV1), and emerging (VVC) video codecs and pro-vide preliminary insights into their applicability in thehealthcare domain.

Experimental Setup The dataset used in this series of experiments consisted of 10atherosclerotic plaque ultrasound videos with a video resolu-tion of 560x448 at 40 frames per second, with a duration of10 seconds, and yuv420p raw format (see Figure 1). Selectedquantization parameters values for constant quality

encoding, typical in video compression performance com-parison studies, were {27, 35, 46, 55} for AV1, VP9 and {22,27, 32, 37} for VVC, HM, and x265, to enable a fair compar-ison and a wide range of representative bandwidths given thevideo characteristics. Random access settings involved anintra update every 32 frames, aligned with the specifiedgroup of pictures (GOP) level. Default preset parametersavailable in every video coded were set at RandomAccessfor VVC and HM, --good and –best for AV1 and VP9,respectively, and –placebo for x265.

Performance EvaluationPreliminary results depicted in Figure 2 and Table 1 showthat VVC already appears to outperform AV1, despite VVC


Figure 1: Original (uncompressed) ultrasound video image example

of the internal carotid artery (ICA). The straight white line

delineates the atherosclerotic plaque causing stenosis. Video

resolution: 560 × 448, frame rate: 40 fps.

Figure 2: Video coding standards comparison. Rate-distortion curves

depicting Peak Signal to Noise Ratio (PSNR) vs log (Bitrate) using

mean values of the ten 560x416@40 frames/s ultrasound videos for

all investigated QP values. VVC outperforms all rival video codecs.

Table 1: Performance comparison of video coding standards in terms

of overall bitrate gains. Savings were computed using the BD-Rate

algorithm. VVC outperformed all other video codecs.

Bitrate savings relative to:

Encoding HM (HEVC) X265 (HEVC) AVI VP9

VVC 23% 33% 33% 54%

HM (HEVC) 13% 18% 40%

X265 (HEVC) 4% 30%

AVI 25%


being in the early stages of the development process. For thisparticular experimental setup, bitrate demands reductions of36 % were recorded for equivalent quality using the BD-RATE algorithm. Interestingly, both HEVC instantiations,namely HM (HEVC Test Model) and x265, demonstratedhigher coding efficiency than AV1, with AV1 entailing addi-tional bitrate requirements of 18 % and 4 %, respectively. Itis important to note, however, that this finding is largely dueto the limited video resolution of the investigated CCAvideos. Ongoing experimentation shows that as video resolu-tion increases, AV1 tends to outperform both HM and x265,which is consistent with the published literature. While notobsolete, VP9 is mostly used for benchmarking purposes inthis study, significantly lacking in compression efficiencycompared to all rival codecs. On a different note, besidesx265 and VP9 that involve production level software opti-mization and hence can qualify for real-time performance,VVC, AV1, and HM incur long compression times, giventhat their principal intended usage is to validate their com-pression efficiency rather than for use in real-time videostreaming applications.

Undoubtedly, more comprehensive experiments are neededto deduct the best video coding software, investigating dif-ferent medical video modalities and higher video resolutions(high and ultra-high definition, and beyond). Studies on gen-eral-purpose videos in the literature revealed that encodingperformance is largely affected by content and video size,and hence safe conclusions can only be drawn once experi-ments extend across the video resolution ladder and involvea sufficient number of medical video modalities. Such astudy is already underway for emergency scenery andechocardiogram videos within the context of the AdaptiveVideo Control for Real-time Mobile Health Systems andServices (ACTRESS) project.

This work was co-funded by the European RegionalDevelopment Fund and the Republic of Cyprus through theResearch and Innovation Foundation (Project: POST-DOC/0916/0023); Acronym: ACTRESS.

Links:

[L1] https://aomedia.org/[L2] https://jvet.hhi.fraunhofer.de/

References:

[1] Cisco, V. N. I.: “Cisco Visual Networking Index: Fore-cast and Trends, 2017–2022”, White Paper, 2018.

[2] A. S. Panayides, M. S. Pattichis, C. P. Loizou, et al.:“An Effective Ultrasound Video Communication Sys-tem Using Despeckle Filtering and HEVC”, in IEEE JBiomed Health Inform, vol. 19, no. 2, pp. 668-676,2015.

[3] Z. C. Antoniou, A. S. Panayides, M. Pantzaris, et al.:“Real-Time Adaptation to Time-Varying Constraints forMedical Video Communications”, in IEEE J BiomedHealth Inform, vol. 22, no. 4, pp. 1177-1188, July 2018.

Please contact:

Andreas S. Panayides Sigint Solutions and University of Cyprus, [email protected]; [email protected]

The vR4REhAB Interreg

Project: Five hackathons

for virtual and Augmented

Rehabilitation

by Daniele Spoladore, Sonia Lorini and Marco Sacco(STIIMA-CNR, EUROVR)

Virtual and augmented reality technologies can support

the process of clinical rehabilitation, making therapy

more engaging, challenging and measurable for people

with disabilities or injury, encouraging them to continue

their rehabilitation outside of the clinical environment

[1]. Although these technologies are used in some

rehabilitation-related fields [2, 3], their extensive

application in the clinical field is still limited.

The European Project VR4REHAB aims to foster the adop-tion of novel virtual reality (VR) applications for rehabilita-tion and health care. The project, a collaboration betweenseven European partners (Sint Maartenskliniek, EuropeanAssociation of Virtual Reality and Augmented Reality(EUROVR), St. Mauritius Therapieklinik, TeessideUniversity, Royal Free London NHS Trust, Université deLille 1 - Sciences et Technologies and Games for HealthEurope), aims to create innovative VR technologies for reha-bilitation clinics. Indeed, virtual and augmented reality canhelp patients recover at home, with or without therapists, byfollowing virtual instructions while being monitored by a setof sensors. Moreover, the virtual and augmented applicationcan calibrate the difficulty of tasks according to the patient’sability, thus reducing boredom associated with too-easy tasksand the frustration that too-difficult tasks would generate.This can result in better engagement rates.

45

Figure 1: UK Hackathon (4 - 5 July 2018) & French Hackathon (14

- 15 June 2018). Participants of the event experiencing VR tools and

pitching their ideas to an international audience.

In Memory

of Peter Inzelt

Dr. Péter Inzelt, former director of theInstitute for Computer Science andControl (SZTAKI) passed away at age75, on the 23rd of June, 2019.

Péter worked for the predecessor ofSZTAKI from 1968 - first as researcherat Department of Control of ContinuousProcesses, then as head of the scientificdepartment from 1981. He becamechief financial officer of SZTAKI in1987, and director in 1993. Péter led theInstitute successfully for 21 years, up toage 70. During his long directorateperiod the Institute strengthened its pro-fessional reputation both at home and inthe international scientific community.

His leadership style was alwaysstrongly characterised by managerialinnovation, fairness in financial affairs,an anti-corruption attitude and socialsensitivity. He introduced a new moti-vation system, launched a voluntary

pension fund for the SZTAKIemployees and their family members,purchased holiday apartments for theInstitute, thus ensuring the summerrecreation of many families and sup-ported sporting activities within andoutside the Institute. He also providedfinancial assistance to families suf-fering from health problems. And thislist is far from complete.

Péter signed SZTAKI’s ERCIM mem-bership in November 1994 during the

ERCIM meetings in Barcelona and wasSZTAKI’s representative until hisretirement. He played an active role atthe Board of Directors where he washighly respected. He always consideredERCIM as one of the most importantorganisations, if not the most important,that SZTAKI joined.

Last, I would like to cite his own wordstaken from his book published for the50-year anniversary of SZTAKI:

“As for me, it was a pleasure to act asdirector here for 21 years, up to thelimit of the legally possible age and myendeavour was always to serve theInstitute in the best way. I hope some ofmy colleagues share this view. I wishmuch success to my successor, to theInstitute, all its managers andemployees.”

SZTAKI considers Péter Inzelt one ofour own. His memory will live on.

László Monostori, Director, SZTAKI

46 ERCIM NEWS 119 October 2019

In 2018, the VR4REHAB project ran five hackathons acrossEurope (in The Netherlands, Germany, France, UnitedKingdom and Belgium), gathering talented youngsters,entrepreneurs, patients, rehabilitation personnel and tech-nical specialists, to develop new ideas about rehabilitationwith virtual and augmented reality. Specifically, the projectfostered the development of new solutions to tackle five clin-ical themes: (i) pain management; (ii) engagement, andimmersion to promote treatment adherence, (iii) behaviouraland cognitive training in children with brain injury, 4) lowerlimbs and mobility; (iv) training of upper limb movements.

Following the hackathons, the best ideas enter a develop-ment phase, the Game Jams, during which the selected con-cepts are developed in prototypes with the help of companiesand ICT professionals; the developed prototypes will betested in clinical trials throughout 2019. Finally, in the lastphase of the project, the Challenges, the prototypes willbecome real applications: the developers, supported by pro-fessionals, will develop business plans to put their applica-tion on the market and to reach their target groups.

VR4REHAB’s five hackathons proved that the hackathonidea was a success: more than 270 people between 18 and 45years old, such as young innovators, developers, researchers,patients, and health care professionals, participated in theevent.

The ideas generated during the hackathons will be stored andpublished into an online library: a digital platform in which

Péter Inzelt (right) and Alain Bensoussan

(left) signing SZTAKI’s ERCIM membership

in 1994.

the concepts will be available for further development.Moreover, the online library will be made available to healthprofessionals, who can provide feedback or helping thedevelopers in fine-tuning their applications. This platformwill be integrated into EuroVRs’ website by the end of theproject.

This project was financed by the EUx InterReg NWE pro-gramme.

References:

[1] S. Arlati, et al.: “A Social Virtual Reality-Based Appli-cation for the Physical and Cognitive Training of theElderly at Home”, Sensors 19.2 (2019): 261.

[2] D. Spoladore, et al.: “Semantic and Virtual Reality-enhanced configuration of domestic environments: TheSmart Home Simulator”, Mobile Information Systems2017 (2017).

[3] D. Baldassini, et al.: “Customization of domestic envi-ronment and physical training supported by virtual reali-ty and semantic technologies: A use-case” 2017 IEEE3rd International Forum on Research and Technologiesfor Society and Industry (RTSI). IEEE, 2017.

Please contact:

Sonia Lorini, European Association of Virtual and Aug-mented Reality, Belgium [email protected]://www.nweurope.eu/VR4REHABhttps://www.facebook.com/VR4RehabProject

ANNOuNCEMENTS

hORIZON 2020 Project Management

A European project can be a richly rewarding tool for pushing your research orinnovation activities to the state-of-the-art and beyond. Through ERCIM, ourmember institutes have participated in more than 90 projects funded by theEuropean Commission in the ICT domain, by carrying out joint research activi-ties while the ERCIM Office successfully manages the complexity of theproject administration, finances and outreach. The ERCIM Office has recog-nized expertise in a full range of services, including identification of fundingopportunities, recruitment of project partners, proposal writing and projectnegotiation, contractual and consortium management, communications and sys-tems support, organization of events, from team meetings to large-scale work-shops and conferences, support for the dissemination of results.

How does it work in practice?

Contact the ERCIM Office to present your project idea and a panel of expertswill review your idea and provide recommendations. If the ERCIM Officeexpresses its interest to participate, it will assist the project consortium either asproject coordinator or project partner.

Please contact:

Peter Kunz, ERCIM Office, [email protected]

Call for Proposals

dagstuhl Seminars

and Perspectives

Workshops

Schloss Dagstuhl – Leibniz-Zentrum für Informatik is accepting proposals for

scientific seminars/workshops in all areas of computer science, in particular

also in connection with other fields.

If accepted the event will be hosted in the seclusion of Dagstuhl’s well known,own, dedicated facilities in Wadern on the western fringe of Germany. Moreover,the Dagstuhl office will assume most of the organisational/ administrative work,and the Dagstuhl scientific staff will support the organizers in preparing, running,and documenting the event. Thanks to subsidies the costs are very low for partici-pants.

Dagstuhl events are typically proposed by a group of three to four outstandingresearchers of different affiliations. This organizer team should represent a range ofresearch communities and reflect Dagstuhl’s international orientation. More infor-mation, in particular, details about event form and setup as well as the proposalform and the proposing process can be found on

https://www.dagstuhl.de/dsproposal

Schloss Dagstuhl – Leibniz-Zentrum für Informatik is funded by the German fed-eral and state government. It pursues a mission of furthering world class researchin computer science by facilitating communication and interaction betweenresearchers.

Important Dates• Proposal submission: October 15 to November 1, 2019• Notification: February 2020• Seminar dates: Between September 2020 and August 2021 (tentative).


Editorial Information

ERCIM News is the magazine of ERCIM. Published quarterly, it

reports on joint actions of the ERCIM partners, and aims to reflect

the contribution made by ERCIM to the European Community in

Information Technology and Applied Mathematics. Through short

articles and news items, it provides a forum for the exchange of

information between the institutes and also with the wider scientific

community. This issue has a circulation of about 6,000 printed

copies and is also available online.

ERCIM News is published by ERCIM EEIG

BP 93, F-06902 Sophia Antipolis Cedex, France

+33 4 9238 5010, [email protected]

Director: Philipp Hoschka, ISSN 0926-4981

Contributions

Contributions should be submitted to the local editor of your

country

Copyright notice

All authors, as identified in each article, retain copyright of their

work. ERCIM News is licensed under a Creative Commons

Attribution 4.0 International License (CC-BY).

Advertising

For current advertising rates and conditions, see

https://ercim-news.ercim.eu/ or contact [email protected]

ERCIM News online edition: ercim-news.ercim.eu/

Next issue:

January 2020, Special theme: Educational Technology

Subscription

Subscribe to ERCIM News by sending an email to

[email protected]

Editorial Board:

Central editor:

Peter Kunz, ERCIM office ([email protected])

Local Editors:

Austria: Erwin Schoitsch ([email protected])

Cyprus: Georgia Kapitsaki ([email protected]

France: Christine Azevedo Coste ([email protected])

Germany: Alexander Nouak ([email protected])

Greece: Lida Harami ([email protected]),

Athanasios Kalogeras ([email protected])

Hungary: Andras Benczur ([email protected])

Italy: Maurice ter Beek ([email protected])

Luxembourg: Thomas Tamisier ([email protected])

Norway: Monica Divitini, ([email protected]),

Poland: Hung Son Nguyen ([email protected])

Portugal: José Borbinha ([email protected])

Sweden: Maria Rudenschöld ([email protected])

Switzerland: Harry Rudin ([email protected])

The Netherlands: Annette Kik ([email protected])

W3C: Marie-Claire Forgue ([email protected])

Cover illustrtation: source: Shutterstock.

ERCIM is the European host of the World Wide Web Consortium.

Institut National de Recherche en Informatique et en AutomatiqueB.P. 105, F-78153 Le Chesnay, Francewww.inria.fr

VTT Technical Research Centre of Finland LtdPO Box 1000FIN-02044 VTT, Finlandwww.vttresearch.com

SBA Research gGmbHFloragasse 7, 1040 Wien, Austriawww.sba-research.org/

Norwegian University of Science and Technology Faculty of Information Technology, Mathematics and Electri-cal Engineering, N 7491 Trondheim, Norwayhttp://www.ntnu.no/

Universty of WarsawFaculty of Mathematics, Informatics and MechanicsBanacha 2, 02-097 Warsaw, Polandwww.mimuw.edu.pl/

Consiglio Nazionale delle RicercheArea della Ricerca CNR di PisaVia G. Moruzzi 1, 56124 Pisa, Italywww.iit.cnr.it

Centrum Wiskunde & InformaticaScience Park 123, NL-1098 XG Amsterdam, The Netherlandswww.cwi.nl

Foundation for Research and Technology – HellasInstitute of Computer ScienceP.O. Box 1385, GR-71110 Heraklion, Crete, Greecewww.ics.forth.grFORTH

Fonds National de la Recherche6, rue Antoine de Saint-Exupéry, B.P. 1777L-1017 Luxembourg-Kirchbergwww.fnr.lu

Fraunhofer ICT GroupAnna-Louisa-Karsch-Str. 210178 Berlin, Germanywww.iuk.fraunhofer.de

RISE SICSBox 1263, SE-164 29 Kista, Swedenhttp://www.sics.se/

Magyar Tudományos AkadémiaSzámítástechnikai és Automatizálási Kutató IntézetP.O. Box 63, H-1518 Budapest, Hungarywww.sztaki.hu/

TNOPO Box 968292509 JE DEN HAAGwww.tno.nl

University of CyprusP.O. Box 205371678 Nicosia, Cypruswww.cs.ucy.ac.cy/

ERCIM – the European Research Consortium for Informatics and Mathematics is an organisa-

tion dedicated to the advancement of European research and development in information tech-

nology and applied mathematics. Its member institutions aim to foster collaborative work with-

in the European research community and to increase co-operation with European industry.

INESCc/o INESC Porto, Campus da FEUP, Rua Dr. Roberto Frias, nº 378,4200-465 Porto, Portugal www.inesc.pt

I.S.I. – Industrial Systems InstitutePatras Science Park buildingPlatani, Patras, Greece, GR-26504 www.isi.gr

Subscribe to ERCIM News and order back copies at https://ercim-news.ercim.eu/

SIMULAPO Box 1341325 Lysaker, Norwaywww.simula.no

special theme: smart things · 2019. 10. 21. · by gclaudi aeck er t, h d of h f raunhofe i ns t...

Documents